Mit

A Literature Study of the Arching Effect
by
Hsien-Jen Tien
B.Sc. Civil Engineering (1990)

National Taiwan University
Submitted to the Department of Civil and Environmental

Engineering in Partial Fulfillment of the Requirements
for the Degree of
Master of Science
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
February 1996
1996 Hsien-Jen Tien

All rights reserved
The author hereby grants to MIT permission to reproduce and to distribute publicly paper
and electronic copies of this thesis document inwhole or inpart.
Signature of Author .......... .. e=..........
D blt of Civil and Environmental Engineering

February 29, 1996
Certified by ............................. ..................

Prof. Herbert H. Einstein
Thesis Supervisor
Accepted by .............................. ... r
MASSACHUSETTS INSTITUTE PrOf. Joseph M. Sussman

OF TECHNOLOGY Departmental Committee on Graduate Studies
JUN 0 5 1996
LIBRARIES
0
A LITERATURE STUDY OF THE ARCHING EFFECT
by
HSIEN-JEN TIEN
Submitted to the Department of Civil and Environmental Engineering

on Feb. 29, 1996 in partial fulfillment of the requirements for the
Degree of Master of Science in Civil and Environmental Engineering
ABSTRACT
If a portion of an otherwise rigid support of a granular mass (e.g. sand) yields, the adjoin-
ing particles move with respect to the remainder of the granular mass. This movement is
resisted by shearing stresses which reduce the pressure on the yielding portion of the sup-
port while increasing the pressure on the adjacent rigid zones. This phenomenon is called
the "arching effect".
The arching effect can be found widely in natural terrain as well as in man-made construc-
tion. It is an important topic for the geotechnical engineers and researchers to compre-
hend and recognize in their work and studies. A review of research related to the arching
effect is presented. The review starts with the classical arching theories including Ter-
zaghi's study (1936, 1943), then extends to the analytical methods, numerical analyses,
empirical approaches, and experimental investigations. A new experimental approach with
the application of photoelastic materials is introduced at the end of this review.
Thesis Supervisor: Dr. Herbert H. Einstein

Title: Professor of Civil and Environmental Engineering
ACKNOWLEDGMENTS
The author extends his sincere gratitude to Professor Herbert Einstein for his in-
sightful help and encouragement throughout the development of this thesis. Thanks are
owed to Professor Samuel Paikowsky of UMASS/Lowell for his precious advice and sup-
port.
The facilities of the Department of Civil and Environmental Engineering in

UMASS/Lowell were used in this research work. The presented research was supported
by the National Science Foundation under grant No. MSS-9358090. The use of these
facilities and the assistance of the NSF are greatly appreciated.
Thanks to the entire Civil Engineering faculty in MIT with special mention to Dr.
John Germaine for his help in my graduate study. Also thanks to Shun-Min Lee and Lucy
Jen in MIT for their friendship and Edward Hajduk in UMASS/Lowell for his help in this
thesis.
Finally, I want to thank my parents who always provided support, encouragement,

and understanding. Without your help, this cannot be possible. I love you.
TABLE OF CONT
Page
ABSTRACT
ACKNOWLEDGMENTS
TABLE OF CONTENTS
LIST OF FIGURES
LIST OF TABLES
CHAPTER 1 INTRODUCTION: DEFINITION A

ARCHING
1.1 General Background
1.2 Definition of Arching
1.3 Brief History of Arching Studies
1.4 Scope of Study
CHAPTER 2 RELEVANCE AND APPLICATIOI

GEOTECHNICAL ENGINEERING 27
2.1 Overview 27
2.2 Geological Structures 27
2.2.1 Review of Karst Terrain and D 27
2.2.2 Relevance to Arching Effect 30
2.3 Man-made Structures 31
2.3.1 Pile Plugging Problems 31
CHAPTER 3 CLASSICAL ARCHING THEORIE

3.1 General
3.2 Terzaghi's Investigations of Arching
3.2.1 Background
3.2.2 Terzaghi's Trap Door Experim
I
3.2.3 Terzaghi's Arching Theory

3.2.4 Application in Tunnel Design
3.3 Loads on Buried Conduits
3.3.1 Background
3.3.2 Assumptions
3.3.3 Design of Buried Conduits
3.3.4 Discussion of the Coefficient of Lateral Stress (K)
3.4 Silo Theory
3.5 Ground Arch Approach and Ground Dome Approach
3.5.1 Background
3.5.2 Ground Arch/Dome Approach
CHAPTER 4 ANALYTICAL APPROACHES

4.1 Introduction
4.2 Continuum Approaches Using Elasticity Theory
4.3 Continuum Approaches Using Plasticity Theory
4.4 Discontinuum Approaches
4.5 Numerical Methods
CHAPT TER 5 EMPIRICAL METHODS 112

5.1 General 112
5.2 Classification of Empirical Methods 112
5.3 Methods Considering the Overburden Depth 113
5.3.1 Bierbiiumer's Theory 113
5.3.2 Balla's Theory 116
5.3.3 Terzaghi's Recommendations for Load on Tunnel Supports 119
5.4 Methods Neglecting the Overburden Depth 122
5.4.1 Kommerell's Theory 123
5.5 Discussion 125
E
CHAPTER 6 EXPERIMENTAL INVESTIGATIONS AND

PHOTOELASTICITY METHODS 135
6.1 General 135
6.2 Experiments by McNulty 135
6.3 Experiments by Ladanyi and Hoyaux 138
6.4 Experiments by Harris 140
6.5 Arching in Granular Soil 141
6.5.1 Experimental Setup 141
6.5.2 Test Results 142
6.5.3 Comparison of Plasticity Theory Approach and Test Results 144
6.6 Centrifuge Modeling of Jointed Rock 146
6.6.1 Introduction 146
6.6.2 Trap Door Experiments Using Centrifuge Modeling 147
6.6.3 Experimental Results 148
6.7 Photoelastic Methods 149
6.7.1 Introduction 149
6.7.2 Photoelastic Study in Embedded Structural Elements 150
6.7.3 Photoelastic Study in Direct Shear of Granular Material 152
CHAPTER 7 SUMMARY, CONCLUSIONS AND RECOMMENDATIONS 185

7.1 Summary and Conclusions 185
7.2 Recommendations for Future Research 188
REFERENCES 190
E
LIST OF FIGURES
Page
1.1 Stress Distribution in the Soil Above a Yielding Base
(Bjerrum et. al., 1972; Revised by Evans, 1984) 23
1.2 Active Arching (Evans, 1984) 24
1.3 Passive Arching (Evans, 1984) 25
1.4 Typical deformation and stress distributions around a rectangular
structure with flexible sides (Evans, 1984) 26
2.1 Subsidence Doline in Alluvium Over Limestone in May River
Dam near Konya, Turkey (Jennings, 1971) 34
2.2 Different Types of Dolines (Jennings, 1971) 35
2.3 Failure Processes of the Cavity Structure
(Benson & LaFountain, 1984) 36
2.4 Penetration States of the Open Pipe Pile (Paikowsky, 1989) 37
2.5 The "Arching Approach" to the State of Stress in the Inner Soil
Plug (Paikowsky, 1989) 38
2.6 Distortion of Soil Due to Sampler Plugging - Passive Arching
(Paikowsky, 1989) 39
3.1 Terzaghi's Experimental Set-up (Terzaghi, 1936) 65
3.2 Terzaghi's Experimental Results: Vertical Force on Trap Door
vs. Displacement of Trap Door
(Terzaghi, 1936; Revised by Evans, 1984) 65
3.3 Terzaghi's Experimental Results: Vertical and Horizontal Stresses
in Soil Body vs. Depth (Terzaghi, 1936; Revised by Evans, 1984) 66
3.4 Terzaghi's Experimental Results: Coefficient of Lateral Earth
Pressure (K) vs. Depth (Terzaghi, 1936; Revised by Evans, 1984) 66
3.5 Yielding in Soil Caused by Downward Movement of a Long
Narrow Section (Terzaghi, 1943) 67
3.6 Free Body Diagram for a Slice of Soil in the Yielding Zone
(Terzaghi, 1943) 67
3.7 (a) Flow of Soil Toward Shallow Tunnel When Yielding Happened
in the Soil Body, and (b) Vertical Stress Profile in Soil Located
above the Tunnel (Terzaghi, 1943) 68
3.8 (a) Yielding Zone in Soil When Tunnel Located at Great Depth,
and (b)Vertical Stress Profile in Soil Located above the Tunnel
(Terzaghi, 1943) 68
3.9 Various Classes of Conduit Installations
(Spangler & Handy, 1973) 69
3.10 Arching Effect inUnderground Conduits
(Spangler & Handy, 1973) 70
3.11 Free Body Diagram for Ditch Conduit (Spangler &Handy, 1973) 71
3.12 Load Distribution at Level of Top of Pipe
(Spangler &Handy, 1973) 71
3.13 Settlements Which Influence Loads on Positive Projecting
Conduits (Spangler &Handy, 1973) 72
3.14 Comparison of Measured Loads on Rigid Pipe and Flexible
Pipe (Spangler &Handy, 1973) 73
3.15 The Coefficient of Lateral Stress (K) for the Design of
Underground Conduits (Iglesia et. al., 1990) 74
3.16 Free Body Diagram for Silo Theory (Origin form Janssen, 1895;
Picture shown here revised by Evans (1984)) 75
3.17 Free Body Diagram for Nielson's Arching Analysis
(Nielson, 1966) 76
3.18 Deformation of Tubes: Mode Two and Mode Four
(Luscher & Hoeg, 1964) 77
3.19 Soil Arching as A Thrust Rings About the Structure
(Luscher &H6eg, 1964) 78
3.20 Trap Door Test Instrument with Various Roof Shapes
(Getzler et. al., 1968) 78
3.21 Soil Arching as An Arch Above the Structure
(Getzler et. al., 1968) 79
4.1 Boundary Conditions for Soil Mass with Yielding Base
(Finn, 1963) 95
4.2 Typical Distribution of Change in Vertical Stress for Downward
Translating Trap Door (Finn, 1963; Revised by Evan, 1984) 96
4.3 Translating Retaining Wall (Finn, 1963) 96
4.4 Boundary Condition for Chelapati's Analysis of Arching in
Granular Material (Chelapati, 1964) 97
4.5 Chelapati's Technique for Elimination of Tensile Stresses using an
Elastic Solution for Arching
(Chelapati, 1964; Revised by Evans, 1984) 98
4.6 Arching Model in the study of Bjerrum et. al.
(Bjerrum et. al., 1972) 98
4.7 Idealized Model for Trap Door Problems (Evans, 1984) 99
4.8 Plastic Flow Rule (Evans, 1984) 99
4.9 Free Body Diagrams for Active Arching in the Two-Dimensional
Case (Evans, 1984) 100
4.10 Free Body Diagrams for Passive Arching in the Two-Dimensional
Case (Evans, 1984) 101
4.11 Soil Cone for Active Arching Above a Circular Trap Door
(Evans, 1984) 102
4.12 Soil Cone for Passive Arching Above a Circular Trap Door
(Evans, 1984) 103
4.13 Soil Prism for Active Arching Above a Rectangular Trap Door
(Evans, 1984) 104
4.14 Soil Prism for Passive Arching Above a Rectangular Trap Door
(Evans,1984) 105
4.15 Model for Systematic Arching Theory
(Trollope, 1957; Revised by Evans, 1984) 106
4.16 Geometrical Definition of Elliptic Structure (Maeda et. al., 1995) 107
4.17 Homogenization of External Contact Forces with Stress
(Maeda et. al., 1995) 107
4.18 Computational Model of the Finite Difference Analysis
(Getzler, et. al., 1970) 108
4.19 Superposition of Two Different Modes (Getzler, et. al., 1970) 108
4.20 Finite Element Discretization of the Trap Door Problem
(Koutsabeloulis and Griffiths, 1989) 109
4.21 Numerical Analysis Results (Koutsabeloulis and Griffiths, 1989) 110
4.22 Particle Configurations and Velocity Vector Fields
(Sakaguchi and Ozaki, 1992) 111
5.1 Rock Pressure Bulb after Bierbaumer (Sz6chy, 1973) 128
5.2 Assumption Model of Bierbiumer's theory: Maximum Load
(Sz6chy, 1973) 129
5.3 Assumption Model of Bierbiumer's theory: Minimum Load
(Iglesia et. al., 1990) 129
5.4 Principle of Balla's Theory (Balla, 1963) 130
5.5 Resistance Factor Diagram and Table (Balla, 1963) 130
5.6 Configuration of Ground Arch (Proctor and White, 1946) 131
5.7 Load-Depth Relationship on a Tunnel in Sand or Crushed
Rock (Proctor and White, 1946) 132
5.8 Simplified Model of Load on Tunnel Support
(Proctor and White, 1946) 133
5.9 Shape of Kommerill's Pressure Diagram (Szechy, 1973) 134
6.1 Layout of McNulty's Experiments (McNulty, 1965) 156
6.2 Active and Passive Arching Curves of Sand 2 (McNulty, 1965) 157
6.3 Influence of Soil Properties and Overburden Depth on Active
Arching (McNulty, 1965) 158
6.4 Ladanyi and Hoyaux's Experimental Setup
(Ladanyi and Hoyaux, 1969) 158
6.5 View of the displacement trajectories for (a) downward, and
(b) upward movement of the structure
6.6 Deformed Square Grid after a Large Downward Movement of
the Trap Door (Ladanyi and Hoyaux, 1969) 160
6.7 Normalized Pressure versus Settlement Curves of Active Arching
6.8 Comparison of Measured and Calculated Vertical Pressures for a
downward moving trap door at different depths of burial
6.9 Apparatus Used in Harris' Experiments (Harris, 1974) 162
6.10 Vertical Stress Distribution from Harris' Experimental
Investigation (Harris, 1974) 162
6.11 Test Apparatus with a Circular Trap Door (Evans, 1984) 163
6.12 Test Apparatus with Rectangular Trap Doors (Evans, 1984) 164
6.13 Variations of Soil Displacement Pattern with Increasing Trap
Door Displacement during an Active Arching Test (Evans, 1984) 165
6.14 Active Arching Test Results (Evans, 1984) 166
6.15 Typical Patterns of Soil Deformation During Passive Arching
Tests (Evans, 1984) 167
6.16 Passive Arching Test Results (Evans, 1984) 168
6.17 Soil Deformations Above Trap Doors Lowered in Sequence,
Active Arching Case (Evans, 1984) 169
6.18 Soil Deformations Above Trap Doors Raised in Sequence,
Passive Arching Case (Evans, 1984) 170
6.19 Comparison of Experimental Results with Those Predicted for
Plane Strain Active Arching at the Maximum Arching State
(Evans, 1984) 171
Plane Strain Active Arching at the Ultimate Arching State
(Evans, 1984) 172
Plane Strain Passive Arching at the Maximum Arching State
(Evans, 1984) 173
Plane Strain Passive Arching at the Ultimate Arching State
(Evans, 1984) 174
6.23 The Trap Door Concept (Iglesia et. al., 1990) 175
6.24 Centrifuge Trap Door Apparatus (Iglesia et. al., 1990) 176
6.25 Test Results Using Coarse Sand of Varying Depths on 1" Door,
Acceleration = 80 g (Iglesia et. al., 1990) 177
6.26 Test Results to Determine Effects of g-Level
(Iglesia et. al., 1990) 178
6.27 Final Configuration of Direct Stack 1/2" Aluminum Rods on
2" Door (Iglesia et. al., 1990) 179
6.28 Sketch of Riley's Model Showing the Location of the Hole, and
the Symmetric Free Field Point (Riley, 1964) 180
6.29 Microflash Photographs Showing the Fringe Order Distribution
around the Boundary of the Hole at 1050 Microseconds after
Detonation of the Explosive Charge (Riley, 1964) 181
6.30 Static and Dynamic Stress Distributions on the Hole Boundary
1050 Microseconds after the Explosion (Riley, 1964) 181
6.31 Plot of Riley's Model Showing the Location of the Hole or
Inclusion and the Loaded Edge of the Plate (Riley, 1964) 182
6.32 Layout of the Shear Box and Sample (Paikowsky et. al., 1995) 183
6.33 Plot of the Shear Box and Sample after One Particle Diameter
Shear at the Bottom (Paikowsky et. al., 1995) 184
LIST OF TABLES
Page
5.1 Overburden Load (in feet) in Sand and in Blocky and Seamy Rock
(Proctor and White, 1946) 127
6.1 Summary of Plane Strain Active Arching Formulae (Evans, 1984) 154
6.2 Summary of Plane Strain Passive Arching Formulae (Evans, 1984) 155
Chapter ONE
Introduction: Definition and Description of
Arching
1.1 General Background
Arching effect is one of the most universal phenomena encountered in soils both in
the field and in the laboratory (Terzaghi, 1943). This effect is most recognized in under-
ground structures, for example, underground conduits. Underground openings can be
built utilizing the arching action to account for the reduction in the overburden pressure.
The nature of the stress redistribution influences the load that reaches the structure,
whether it is from overburden soil, surface surcharge, or lateral earth pressure. The soil
medium adjacent to the underground opening can increase the structure's load-carrying
ability compared to an identical unburied structure. A more detailed definition of arching
is presented in the following section. Arching effect can also be found in the natural land-
scape. For example, the arching phenomenon in karst terrain can be observed easily and
widely.
Since arching affects geotechnical engineering, people have tried to understand its
mechanism for decades. Much research has been done in this area, including theoretical
derivations, analytical methods, numerical analyses, and experimental investigations. Re-
searchers also applied the arching theory to practical engineering problems, for example,
the soil plug problem (Paikowsky, 1989) and the sheet pile design (Rowe, 1952).
This literature study reviews past research related to the arching effect and at-
tempts to sort the available information for future research use. The following five cate-
gories are covered in this study: (1) Relevance and Application of Arching in Soil Mechan-
ics, (2) Classical Arching Theories, (3) Analytical Approaches, (4) Empirical Methods,
and (5) Experimental Investigation. A discussion and relevant comments are provided,
related to the interaction between the various categories. Conclusions and discussion con-
cerning the existing arching studies as well as suggestions for future work are presented in
Chapter 7.
1.2 Definition of Arching
Arching can be best described as a transfer of forces between a yielding mass of
geomaterial and adjoining stationary members. A redistribution of stresses in the soil body
takes place. The shearing resistance tends to keep the yielding mass in its original position
resulting in a change of the pressure on both of the yielding part's support and the adjoin-
ing part of soil (Terzaghi, 1943). If the yielding part moves downward, the shear resis-
tance will act upward and reduce the stress at the base of the yielding mass (Figure 1.1).
On the contrary, if the yielding part moves upward, the shear resistance will act downward
to impede its movement and cause increase of stress at the support of the yielding part.
Depending upon relative stiffnesses in the ground mass, arching can either be ac-
tive or passive. Active arching occurs when the structure is more compressible than the
surrounding soil, as illustrated in Figure 1.2 (a). When the system is subjected to loads,
the resulting stress distribution across locations of equal initial elevation (Plane AA and
Plane BB) is similar to that shown in Figure 1.2 (b), where the stresses on the structure
are less than those on the adjacent ground. If the structure deforms uniformly on Plane
AA and BB, the stresses on it tend to be lower toward the edges due to mobilized shear
stresses in the soil.
In passive arching, the soil is more compressible than the structure as illustrated in
Figure 1.3 (a) As a result, the soil undergoes large displacements, mobilizing shear
stresses which increase the total pressure on the structure while decreasing the pressure in
the adjacent ground. Assuming the structural deformations are uniform, the stresses are
highest at the edges and lowest at the centerline. The stress distributions for the passive
case at Plane AA and BB are shown in Figure 1.3 (b).
If the soil medium and the structure have the same constitutive properties (i.e. load
vs. deformation relationship), the stress along a plane (like Plane AA and BB shown in
former figures) will be uniform. The stress along the vertical direction will be linear and
increasing with depth (geostatic stresses) as no arching would be presented in this case.
This condition is highly unlikely to be found in natural or man-made environments due to
the differences in the mechanical properties of geomaterials (like soils or rocks) and
structure components (like steel or concrete).
Underground structures normally do not have uniform deformation resulting in
stress distributions more complex than those discussed above (Figure 1.2 & 1.3). An ex-
ample is presented in Figure 1.4. The horizontal faces (like Plane AA) and vertical faces
(like Plane BB) are more flexible towards the centers of the spans, resulting in the defor-
mation patterns shown in Figure 1.4. The horizontal and vertical stress distributions are
also shown in this figure suggesting that the faces of the
and passive arching at the same time.
1.3 Brief History of Arching Studies
The phenomenon of arching has been recognized
generally been sporadic, usually directed toward a parl
point in time. This section summarizes the developments
Arching is present in many geotechnical problem
recognized and investigated in a non-geotechnical co
French military engineers were asked to design magazin
that the base of the silo only supported a fraction of the
and the side walls carried far more load than anticipate
small section of the base were detached and lowered, t
experienced was independent of the height of material ir
that an "arch" had formed above this displaced section.
edge of the behavior within magazine silos was utilized
approach to the design of silos for grain and other parl
ory").
Around 1910 considerable land drainage project
of the United States. Engineers found that many of the
structural failures subsequent to installation and backfi
Anson Marston (1930) performed extensive research a

loads on underground conduits/pipes, finding that the loads may vary between a small
fraction of the overburden and several times the overburden. Different loads on the con-
duits depended upon the conduit's flexibility and the installation procedure (Spangler,
1964). and this was attributed to arching.
In the 1920's and 1930's the importance of arching around tunnels was recog-
nized. Designers found that the support loads were far less than the overburden and that
considerable savings could be achieved if accurate predictions of load were possible. This
gave rise to empirical relations for tunnel support loading. Some of these relations are still
in use today (Sz6chy, 1973), including Terzaghi's design values for underground structure
support loads under various ground conditions (Terzaghi, 1943; Proctor and White, 1946,
1977). The interest in tunnel support loads also led to experimental and theoretical treat-
ment of the problem (most notable Terzaghi's research in 1936 and 1943).
In the 1950's, the decision to build an interstate highway system in America cre-
ated new interest in the loads on underground conduits. Larger culverts, with fill heights
and culvert loadings greater than ever before, were required. Researchers reviewed and
updated Marston's recommendations in light of experience obtained in the several decades
since his investigations. Particular attention was given the positive effects of load redistri-
bution around flexible culverts, and techniques for reducing the load on a culvert through
specific backfilling procedures (Spangler, 1964; Spangler & Handy, 1973).
The direction of arching-related research shifted once again in the 1960's when the
Defense Department of the United States sponsored considerable research in the area of
soil-structure interaction. Techniques were needed for the design of massive defense fa-
cilities and it was recognized that the arching phenomenon would allow facilities placed
below ground to withstand nuclear attacks during the war, which would destroy any sur-
face facilities (Whitman et. al., 1962, 1963). Most of the research was presented at the
"Symposium on Soil-Structure Interaction" in 1964.
Starting in the 1970's, computer-based techniques have been broadly utilized in the
studies of arching problems. Getzler, et. al. (1970) used the finite difference method to
analyze the arching pressures in an ideal elastic soil model. Rude (1982) utilized a linear
elastic finite element program to predict the behavior of a culvert installed in a laboratory
testing tank. Rude's predictions based on the program had shown good agreement with
experimental results. More recently, individual particle's properties and interparticle rela-
tionships are taken into account in numerical analysis. Sakaguchi and Ozaki (1992) used
the "Discrete Element Method" (DEM) for computer simulations on the formation of
arches plugging flow. They considered the rolling friction effect between particles and got
a good agreement between the simulation outcomes and experimental measurements.
Terzaghi's trap door tests (1936) have been duplicated by several researchers, e.g.
McNulty (1965), Ladanyi and Hoyaux (1969), Harris (1974), Vardoulakis et. al. (1981),
Fricki and Fricker (1983), and Evans (1983). No real advances have been obtained in
these research projects compared to Terzaghi's approach. Lately, some new technologies
were applied in the arching studies. Iglesia, Einstein, and Whitman (1990) used centrifuge
modeling to study scaling issues and arching in geomaterials. However, the stress distri-
bution across the yielding surface and the correct shape of the sliding surfaces are still not
known well. Techniques to measure the distribution of stresses within a soil body are re-
quired. The "Photogrametric Method" (Yoshida et. al., 1993) and the "Tactile Sensing
Method" (Paikowsky & Hajduk, 1996) are the new techniques being investigated recently
for measuring the stress distribution in granular soil. Materials like photoelastic particles
can also be utilized to study the shape of the sliding surfaces related to the arching effect
(Paikowsky et. al., 1996).
1.4 Scope of Study
The scope of this literature review is divided into five major categories:
(1) Relevance andApplication of Arching in Soil Mechanics: (Chapter2)
The first topic is the exploration of the natural terrain or the geologic structures
relevant to the arching effect. The second one is the discussion of man-made construction
which apply the arching theory. Examples for natural geological structures are Karst and
Sinkholes (Beck, 1984) and examples for man-made construction are tunnel construction
(Terzaghi, 1943), buried cylinders and pipes (Marston, 1930; Spangler, 1964), and soil
plugging (Paikowsky, 1989).
(2) ClassicalArching Theories: (Chapter 3)
The classical arching theories in soil mechanics were mostly influenced by Terzaghi
(1936, 1943). Terzaghi conducted the most widely known experimental and theoretical
investigations of arching. The original studies and the associated analytical derivations of
the arching theories are reviewed along with the application of the theories to practical
construction. In addition to Terzaghi's research, the following studies are also reviewed:
(a) Silos theory (Janssen, 1895 & Jakobson, 1958), (b) Marston/Spangler underground
conduit analyses (Marston, 1930 & Spangler, 1964), (c) ground arch/dome approaches,
and (d) Soil Plugging (Paikowsky, 1989),.
(3) AnalyticalApproaches: (Chapter 4)
These approaches include
(a) Continuum Approaches using Elasticity Theory, e.g. Finn (1963),
(b) Continuum Approaches using Plasticity Theory, e.g. Evans (1984),
(c) Discontinuum Approaches, e.g. Trollope (1957, 1963), and
(d) Numerical Methods, e.g. Finite Element Method (Selig, 1975; Rude, 1982;
Einstein, 1980), Finite Difference Method (Getzler, 1970; Chelapati, 1964),
and Distinct Element Method (Sakaguchi and Ozaki, 1993).
(4) EmpiricalMethods: (Chapter5)
Some of the empirical methods utilize simple static analyses while others are en-
tirely founded on the experience of the person proposing it. These methods are separated
into two major areas. One area contains the methods considering the effect of overburden
depth and the other area includes the methods disregarding the effect of overburden depth.
For example, the methods of BierbAumer (Sz6chy, 1973), Balla (1963), and Terzaghi
(1946) considered the effect of overburden depth; the method of Kommerell (Sz6chy,
1973) neglected this effect.

(5) ExperimentalInvestigationandPhotoelasticityMethods: (Chapter6)
All the experimental investigations are model tests which are used to examine the
arching behavior
(a) McNulty's experiment (McNulty, 1965)
(b) Ladanyi and Hoyaux' experiment (Ladanyi & Hoyaux, 1969),
(c) Sandbox trap door experiments (Harris, 1974),
(d) Arching in granular soil (Evans, 1983),
(e) Centrifuge Modeling of Jointed Rock (Iglesia, et. al., 1990), and
(f) The photoelasticity method allows one to study stress distribution around
structures of complex geometry. The previous study has Riley's experiment
(1964) and Paikowsky and Xi's direct shear experiment (1995).
A summary and conclusions for this literature survey will be found in the last
chapter (Chapter 7). Some recommendations for future researches will also be proposed
in that chapter.
04----- nz 1 --
I
l. 2L. of Support
Figure 1.1 Stress Distribution in the Soil Above a Yielding Base (Bjerrum et. al., 1972
Revised by Evans, 1984)
A --- SA
R IIII I - - -- --- -B
(Structure within soil mass, no force present)
U
PS
------
A _ _ft_ .__ .
B------- emu
-
--No .."
-

-
-M
(Pressure Ps applied)
(a) Displacements under pressure Ps when structure is more compressible than surround-
ing soil
nLPS
(b) Stress distribution across Plane AA or BB
Figure 1.2 Active Arching (Evans, 1984)

- - -- A
,-,, B
(Structure within soil mass, no force present)
El
- --
A
PS
(Pressure Ps applied)
(a) Displacements under pressure Ps when structure is less compressible than surrounding
soil
LS
m~ I
i
(b) Stress distribution across Plane AA or BB
Figure 1.3 Passive Arching (Evans, 1984)

A --- A- A
>ution
Vertical stress distribution on Plane AA
Figure 1.4 Typical deformation and stress distributions around a rectangular structure with
flexible sides (Evans, 1984)
Chapter TWO
Relevance and Application of Arching in
Geotechnical Engineering
2.1 Overview
The relevance and application of the arching theory in geotechnical engineering is
reviewed here. In the first part of this chapter, natural terrains and geologic structures are
examined in light of the arching mechanism. Man-made structures as applied to the arch-
ing theory is discussed in the later part. Examples of natural geological structures in
which arching plays a role are Karst and Sinkholes (Beck, 1984). Examples for man-made
structures are tunnels (Terzaghi, 1943), buried cylinders and pipes (Marston, 1930; Span-
gler, 1964), sheet pile designs (Rowe, 1952 & 1955), and pile plugging problems
(Paikowsky, 1989).
2.2 Geological Structures
2.2.1 Review of Karst Terrain and Dolines
Karst is a complex composition of landform and subterranean materials formed by
the dissolution of soluble rocks, e.g. limestone, dolomite, and gypsum. Karst terrain is
quite often mantled with various surface deposits such as weathered debris, terra rosa
soils, loess and alluvium. The most obvious concern of karst terrain is the development of
high permeability as water travels through and dissolves the material. Unique characteris-
tics concerning karst terrain were outlined by LeGrand (1973) as follows:

(1) A general scarcity and poor predictability of groundwater supplies,
(2) A scarcity of surface streams,
(3) Instability of surface streams,
(4) Instability of the ground, and
(5) Leakage of surface reservoirs.
There are many different factors that contribute to the formulation of karst. Not all of the
conditions listed above need to be met in order for karst terrain to develop, nor does the
inclusion of all these factors indispensably mean that karst terrain will emerge. Each karst
terrain has particular qualities that make it unique compared to other karst terrain. An im-
portant factor to remember is that it is a land formation that is in a continuous state of
change.
There are three major rock types that can be found in karst terrain. They include
limestone, dolomite, and evaporites. Limestone exists in most karst terrains due to its
high solubility in water, which makes it susceptible to developing pores. The chemical
composition of the ground water plays an important role in karst composition. In the
event that carbon dioxide is dissolved into the water, the water is transformed into a car-
bonic acid, increasing the dissolution rate. When water circulates through the pores due
to heavy rainfall, there is the influence of both a high energy gradient and a more acidic
solution to deteriorate the soil and the limestone. Figure 2.1 shows a subsidence doline in
alluvium over limestone. Dolomite rock behaves in natural water in an essentially similar
way to that of limestone. However, under normal air and water interface conditions, it is
claimed that dolomite is usually less soluble than limestone (Douglas, 1965). The third
important karst rock is evaporite. Gypsum is an example. It is much more soluble than
either limestone or dolomite (Trombe, 1952).
Dolines (also called Sinks or Sinkholes) are a special geological structure that can
be found broadly in karst terrain. They are usually circular or oval in plan, with depth
varying very much in relation to diameters. Several processes are responsible for sinkhole
formation: surface solution, cave collapse, piping, subsidence, and stream removal of su-
perficial covers. These processes often occur in combination. They are described indi-
vidually as follows (Jennings, 1971):
(1) Solution Dolines: They usually form where structural control such as intersect-
ing joints leads to infiltration of surface water. As more water flows through
the defect, the depression enlarges and leads to even more inflow. Solution
structural control may alter the original geometries. If sediment accumulates,
the floor of a large dolines can be swamp or contain small lakes or ponds. Fig-
ure 2.2 (a) shows the diagram of a solution doline in a jointed rock mass.
(2) Collapse Dolines: They usually result from the collapse of a cave produced by
underground solution. These dolines exhibit collapse features such as steep
walls and an angular shape in plan. The depth-width ratio often is greater than
for the solution dolines. If no further collapse occurs, these dolines eventually
weather and appear more like a solution doline. The configuration of a col-
lapse doline is shown in Figure 2.2 (b).
(3) Subsidence Dolines: They form when sinks form in karst rocks underlying a
superficial deposit or thick residual soil. These dolines can also form through
continuous piping of materials through widening joints or solution pipes. Fig-
ure 2.2 (c) shows the subsidence doline.
(4) Alluvial Stream-sink Dolines: Dolines form in alluvium where streams sink
into underlying karst rock. The processes which create subsidence dolines op-
erate here but additionally the stream provides a good channel for mechanical
removal of the insoluble alluvium. Figure 2.2 (d) shows the alluvial stream-
sink doline.
(5) Subjacent Karst Collapse Doline: Figure 2.2 (e) shows this kind of dolines.
Cave collapse occurs in karst rocks beneath overlying bedrock formations. A
steep-walled, deep doline may form initially (Figure 2.2 (e)) but weathering
will turn them into conical features which may be degraded into still gentler
forms.
2.2.2 Relevance to Arching Effect
Sowers (1984) indicates that the most important problem in residual settlement is
"the collapse of domes within the residual soils and the development of a sinkhole accom-
panied by catastrophic foundation subsidence, often with little warning." The formation of
a sinkhole starts with the creation of a dome cavity that propagates upward until the dome
can no longer support the increased load. For the most part, sinkholes collapse when the
water table is lowered, and the cavern is subject to an increase in effective stress. In other
words, the overburden above the cavern increases such that the dome cavern is unable to
support the increased load.

Figure 2.3 shows the failure processes of the arch/cavity structure (Benson &
LaFountain, 1984). In stage I, the total cavity system and overburden are stable. In stage
II, the stable cavity system has some overburden instability above it. Then a moderate
cavity system and more overburden instability develop in stage III. Finally, the consider-
able instability of the cavity system results in gross overburden instability and small surface
displacement in stage IV After this stage, the whole system fails.
As most sinkhole failures develop as a result of lowering the water table, cohesion
is usually an important factor in determining the failure strength of the cavern roof. For a
cohesionless granular medium, where cohesion plays no role in determining the failure
stress characteristics of the cavern, it is presumed that the arching mechanism that controls
lateral transmission of stresses plays the most significant role. Although arching undoubt-
edly contributes to the strength characteristics of cohesive medium, it's contribution is not
as significant as it is for non-cohesive medium. A need for understanding the arching
problems in granular material exists when one studies the strength of sinkholes comprised
of non-cohesive material.
2.3 Man-made Structures
2.3.1 Pile Plugging Problems
Pile plugging problems are used in this section as an example to explain the rele-
vance and application of arching to man-made structures. The other examples (tunnels
and buried conduits) will be discussed later in Chapter Three.

Pile plugging problems are caused by soil plug behavior. Soil plug behavior hap-
pens during the installation of open pipe piles. At the initial stage of installation, soil en-
ters the pipe pile at the same rate as the pile penetration, and the pile is called "unplugged"
(Figure 2.4 (a)). As the pile is driven to the ground continuously, the inner soil body de-
velops friction on the inner pile wall, which may prevent some of the soil in front of the
opening from entering the pile. When the length of the inner soil cylinder is less than the
penetration depth, the pile is considered as "partially plugged" (Figure 2.4 (b)). If a suffi-
cient friction is developed on the inner pile wall, no more soil can enter the pile and the
pile is called "plugged" (Figure 2.4 (c)). The open ended pile is then assumed to have the
same penetration characteristics of a close ended pile.
Paikowsky (1989) evaluated the inner soil resistance of the pile plugging problems
in his research. The behavior of a soil plug under static loads was analyzed using the "silo
approach" (Jakobson, 1958) as a tool. This approach led to the conclusion that the
"arching effect" controlled the mechanism of load transfer in the silo as well as in an
opening pile. According to Paikowsky's study, two possible relative soil and wall move-
ments can take place: (1) "ACTIVE" (associated with "active arching"), in which the soil
settles with respect to the walls (e.g., the "standard" silo case), and (2) "PASSIVE"
(associated with "passive arching"), in which the walls move downwards with respect to
the soil (e.g., open pile penetration). In this section, the "PASSIVE" case related to the
pile plugging problems will be presented.
Figure 2.5 (a) presents two possible principal stress trajectories in an opening pile.
When the pile moves downwards with respect to the soil, the shear stresses are acting
downwards on the soil and upwards on the inner pile surface. The stresses acting on the
soil and inner pile surface for this case (pile plugging) are shown in Figure 2.5 (b). If the
friction angle between the soil-pile interface is assumed equal to the soil internal friction
angle (<), the stress state at the interface can be described by the Mohr circle diagram
shown in Figure 2.5 (c). After the pole (point "P") is found, the directions of major prin-
cipal stresses can be decided. These directions lead to the concave major principal stress
trajectory as shown in Figure 2.5 (a). This is because inthe condition of pile plugging, the
soil is being pushed upwards in a "passive arching" mode, an arch made of particle con-
tacts is oriented concave downward in the major principal stress direction. The major
principal stress direction at the centerline of the pile is horizontal in this case (Figure 2.5
(a)). A real passive arching case from pile plugging samples is presented in Figure 2.6.
Several concave arches can be observed in this figure. On the other hand, if the shear
stresses act upwards on the soil and downwards on the inner pile surface, all the afore-
mentioned phenomena are reversed. The soil is then in the "active arching" mode, result-
ing in the convex minor principal stress trajectory as shown in Figure 2.5 (a). The major
principal stresses in the soil are then perpendicular to this trajectory and support the load
above it. The detailed mechanism of the active arching (the silo behavior) will be depicted
in the following Chapter.
Because of the passive arching effect, the open pipe piles constitute a strong foun-
dation for land and offshore construction. They are easy to handle and splice, and more
light-weight than the closed end piles.

~ry~~
.4,c;i ,
;~?r~~3~"~:~'~~~e~:~Q ,:~
?~t
47
.- ,,
:XI11 I
T'^ ~'~ :..-~f~c"~~'~*l
i ~it7
. . . .zk., , .- ,
.7 . . -.
iZ
,, . . . '- : , . . . z , . .
Cil~' " .. ~rJ
=e;~-r~ir 5`" 1:~e~M~r~rA~; ,...,
i~f6~j;j~f~r~ 5~f:
Figure 2.1 Subsidence Doline in Alluvium Over Limestone in May River

Dam near Konya,
Turkey (Jennings, 1971)
34
SoLution doLine CoLLapse doLine
(a) (b)
Surface of alluvial plain

Streamsink doine
Subsidence doLine ALLuviaL streamsink doLine Subjacent karst

coLlapse doLine
(c) (d) (e)
Figure 2.2 Different Types of Dolines: (a) Solution Doline, (b)Collapse Doline, (c) Subsi-
dence Doline, (d) Alluvial Stream-sink Dolines, and (e) Subjacent Karst Col-
lapse Doline. (Jennings, 1971)
Figure 2.3 Failure Processes of the Cavity Structure: Stage I - Stable Cavity and Over-
burden System, Stage II - Stable Cavity System with Some Overburden In-
stability, Stage III - Moderate Cavity System with unstable Overburden, Stage
IV - Failure of Cavity System (Benson & LaFountain, 1984)
o
cl:
i,
CD i-I I<
-I i i
.0
2'1
CO
.
a) -
-Jo-r c o m
- . u < I
CC
CD
a-
L.. 0
C14
L..i
o
o -o
CD
C)
~C)
C)
j- II
o
-u
oori]
I,
_.J
1 )
C o
O
U)
So
* oo*
cc~e
0 0
C/
0UUa..
L 0
Bpa
ca n
.=U
oa .0
Lo
oU CO~ 4
ma"
0 +.. "( o '
~Ca (;
C1
I."
9v,
1.*
O ..
Figure
o
"4r C. - .ieA.
l.w 4 I- n4,& (SN O - a a" . 6&W
s, ILSY temo - as"CAAMA - 0
Figure 2.6 Distortion of Soil Due to Sampler Plugging - Passive Arching

(Paikowsky, 1989)
39
Chapter THREE
Classical Arching Theories
3.1 General
Classical arching theories combining experimental observation and theoretical
derivation are introduced and discussed in this chapter. The theories presented here
served as the foundation to the development of many additional arching studies. They
play an important role in exploring the arching effect.
The most widely known experimental and theoretical investigations of arching
were conducted by Terzaghi (1936, 1943). These studies are introduced at the beginning
of this chapter. In addition to his studies, the analyses of loads on buried conduits
(Marston, 1930; Spangler & Handy, 1973), the silo theory (Janssen, 1895; Jakobson,
1958), and ground arch (& dome) approach will be discussed consecutively.
3.2 Terzaghi's Investigations of Arching
3.2.1 Background
In order to improve understanding of the arching phenomenon in general, and
specifically the stress distribution around tunnels, experimental investigations were made
by K. Kienzl in Terzaghi's laboratory in Vienna (Terzaghi, 1936). From these results, the
theories of arching were derived in 1943. Terzaghi (1943) described the arching effect as
"the transfer of pressure from a yielding mass of soil onto adjoining stationary parts" and
the soil is said to "arch over the yielding part of the support". He also combined the
knowledge which was obtained in these studies and information from many underground
construction projects to generate the design values for underground structure support
loads under various ground conditions (Proctor and White, 1946 & 1977). Some of Ter-
zaghi's recommendations are still applied in underground construction practice at present.
3.2.2 Terzaghi's Trap Door Experiment (Terzaehi, 1936)
In Terzaghi's experiment a trap door, which was mounted flush with the base of a
box containing sand, was translated downward while the total load on the door and its
displacement were monitored. Horizontal and vertical stresses at various heights above
the door were indirectly measured using the friction tape method. Terzaghi's experimental
set-up is shown in Figure 3.1. The trap door has a width (2B) of 7.3 cm and a length of
(L) 46.3 cm (out of the figure plane).
Figure 3.2 shows the typical results presented by Terzaghi (1936). The force on
the trap door was normalized by its initial value, i.e. the force acting on the trap door be-
fore the trap door was lowered. In Figure 3.2, the normalized force decreased rapidly as
displacement begun and the minimum value of this normalized force occurred at a dis-
placement of only about 1%of the trap door width. These minimum values were less than
10% of the overburden sand weight and had the tendency to be lower for dense sand (6%)
than for loose sand (9.6%). As displacement continued, the "structure" developed within
the sand was believed to have disintegrated somewhat, causing the load to increase until a
constant value which was still only a small fraction of the overburden (z12.5%). This
constant value was obtained for trap door displacements greater than about 10% of the
door's width. In addition, dense and loose sand showed an ultimate trap door force of the
same magnitude.
Figure 3.3 shows measured vertical and horizontal stresses within the soil profile
above the trap door. The vertical stresses decreased as soon as the trap door was moved
downward. The horizontal stresses increased a little in a part of the soil profile above the
trap door (at the position about H/3) but it decreased to a small value after the augment.
Figure 3.4 shows values for the coefficient of lateral stress (K) obtained from the experi-
ments. For the case of 1% deflection, K was approximately 1.0 directly above the trap
door and increased to about 1.6 at about one trap door width (2B) above the door. At a
distance of 5B above the trap door, K was essential equal to Ko. Terzaghi thought this
result meant that lowering the trap door seemed to have no effect at all on the state of
stress in the sand above this height, i.e. no arching effect outside this range. He suggested
the value of K to be approximately unity, based on experiments and experience (Terzaghi,
1943).
Terzaghi noted that arching does not necessitate the crushing of soil particles to
support the arch formation. It is a temporary circumstance dependent on the shear
stresses in the soil. Vibration is the primary mechanism capable of disturbing the arching
phenomena. Experimental results reported here are cases of "active" arching, which has
upward shear stresses acting at the sides of the soil prism above the trap door. Because of
the upward shear stresses, the normal stresses acting on the trap door were smaller than
the overburden stresses at the top of the trap door.

3.2.3 Terzaghi's Arching Theory (Terzaghi, 1943)
According to the experimental results presented in the last section, Terzaghi pro-
posed a theoretical approach for the arching problems in sand under plane strain condition
(Terzaghi, 1943). He defined the arching effect as the pressure transfer between a yielding
mass of soil and adjoining stationary parts. The relative movement in the soil is opposed
by a shearing resistance within the contact zone of the yielding and stationary masses.
Hence, the pressure transfer is possible through the shearing resistance which plays an
important role in the arching theory.
The real surfaces of sliding, as observed by Terzaghi in 1936, are curved and at the
soil surface their spacing is greater than the width of the yielding strip. The yielding strip
ab at the solid base is presented in Figure 3.5, and the real sliding surfaces are curve ac
and curve db in the same figure. Several assumptions are used in the arching theories
based on the experimental observations. The sliding surfaces are assumed to be vertical.
The vertical sections ae and bf through the outer edges of the yielding strip in Figure 3.5
represent surfaces of sliding. The pressure on the yielding strip is thus equal to the differ-
ence between the weight of the sand located above the strip (ab) and the shear resistance
along the vertical sections. The free body diagram for a slice of soil in the yielding zone
above strip ab is presented in Figure 3.6. In addition to the vertical sliding surface as-
sumption, Terzaghi also assumed that the normal stress is uniform across horizontal sec-
tions and the coefficient of lateral stress (K) is a constant. Cohesion (c) was assumed to
exist along the sliding surfaces. The vertical equilibrium for the free body in Figure 3.6 is:
2Bydz = 2B(ov + dov) - 2Bav + 2cdz + 2ahdz tan 0 (Eq. 3.1)

in which
2B = width of the yielding strip (ab),
z = depth,
y = unit weight of soil,
ov = vertical stress,
Oh = horizontal stress = Kov,
K = the coefficient of lateral stress,
c = cohesion,
= friction angle.
The boundary conditions are Ov = q (surcharge) at z =0. Solving Equation 3.1, leads to:
v=B(y (1- eK ) + q -e (Eq. 3.2)

Ktan4
in which
q = surcharge at the soil surface.
The experimental investigations regarding the state of stress in the sand located
above a yielding strip (see Section 3.2.2) have shown that the arching effect only extends
to a height of 5B. In other words, at elevations of more than 5B above the center line the
lowering of the strip has no effect at all on the state of stress in the sand (Terzaghi, 1936
&1942). Terzaghi assumed therefore that the shear resistance of the sand was active only
on the lower part of the vertical boundaries ae and bf in Figure 3.5. With this assumption,
the upper part of the soil prism (eelfif) is treated as a surcharge q on the lower part
(elabfi). If z, = n1B is the part of prism which acts like surcharge, and z2 = n2B is the part
of prism with shear resistance at the vertical boundaries, then Eq. 3.2 becomes:
v=B(y-B) e-K*n2tan2
B(yv - c) (- e-K 2"')+yBni.e - K ' n2 ' (Eq. 3.3)
Ktan
when we substitute q = yniB and z = n2B to Eq. 3.2. When n2 is very large, the vertical
stress av(o,) is equal to B(y-c/B)/Ktan. This means below certain depth, the vertical stress
on the yielding strip will be a constant.
Finally, several limitations of Terzaghi's arching theory are presented: (1) the ver-
tical stresses on the horizontal yielding surface are assumed to be uniform, (2) the trap
door or yielding strip is assumed rigid, and (3) the assumed sliding surfaces are not true.
For the arching studies in the following sections, researchers have tried to modify these
assumptions in order to make the predictions of their theories closer to the physical condi-
tions.
3.2.4 Application in Tunnel Design
Terzaghi applied the aforementioned theory to tunnel design (Terzaghi, 1943).
The stress state in the soil above the top of a tunnel is similar to the stress state in the soil
above a yielding strip. Terzaghi assumed the soil adjacent to the tunnel yields laterally to-
ward the tunnel during construction. This creates an active earth pressure condition with
the boundaries of the yielding zone inclined at about (450 +/2). The yielding zones at the
sides of the tunnel and the assumed yielding prism (eibiblel) are shown in Figure 3.7 (a).
At the level of the tunnel roof, the width of the yielding strip (2B 1) for a rectangular tunnel
is:
2Bi = 2[Bo + H -tan(45 -- )] (Eq. 3.4)

2
If the tunnel roof is located at a depth D in the ground, the vertical stress on the roof is:
o=I)(1- e - K'ftn ') (Eq. 3.5)

Ktan )
Figure 3.7 (b) shows the vertical stresses in the soil above the tunnel.
If a tunnel is located at a great depth below the surface, the arching effect cannot
extend beyond a certain elevation D1 above the tunnel roof (like z2 in the last section, see
Figure 3.5). Also the soil located above this elevation has a depth D2 (like z1 in Figure
3.5). Figure 3.8 (a) shows the configuration of the tunnel at a great depth. The vertical
stress on the roof is then expressed as:
=B)(yt-
oYv =1- (1)
( e -K.,tn.DI, ) + De -K.nD
K.tan#D)' B (Eq. 3.6)
Ktan (
When D, is very large, the vertical stress ov(.o) will reach a limit value:
ov(-) = (Eq. 3.7)

Ktan
If the tunnel is constructed in sand, then cohesion (c) is equal to 0. However, for safety
reasons, c = 0 is assumed and Eq. 3.7 can be simplified to:
By
ovc()- - B(Eq. 3.8)
Ktan4
Figure 3.8 (b) shows the vertical stress profile at the top of a tunnel at a great depth. The
arching effect only exists within a distance D1 above the tunnel roof (Figure 3.8). There is
no arching effect outside this range, i.e. no shear resistance within D2.
3.3 Loads on Buried Conduits (Spangler and Handy, 1973, 1982)
3.3.1 Background
In 1913, Anson Marston developed a theory to explain the characteristics of a soil
column above a buried conduit. Marston found that the load due to the weight of the soil
above a buried conduit does not fully act on the conduit; part of the weight is undertaken
by the arching action in which load is transferred to the adjacent side material (e.g. soil).
Buried conduits can be grouped according to their installation procedures. The two major
categories are those installed in a ditch excavated through existing soil, i.e. ditch conduit
(Figure 3.9 (a)), and those placed at existing ground level above which an embankment is
subsequently constructed, i.e. projecting conduit. If the top of the structure projects
above the ground surface, it is a "positive" projecting conduit (Figure 3.9 (b)). If it is
placed in a shallow trench and the top lies below the ground surface, it is a "negative"
projecting conduit (Figure 3.9 (c)).
Arching action and the equal and opposite arch support play a tremendously im-
portant role in the development of earth load on a structure. In some cases, such as the
case of a pipe in a trench (a ditch conduit), its effect is favorable; that is, it reduces the
load as compared to the dead weight of the prism of soil lying above the structure (Figure
3.10 (a)). In other cases, such as some installations of culverts under embankments,
arching action may be inverted and the load on the structure may be considerably greater
than the weight of the overlying prism of soil (Figure 3.10 (b)). In this section, the various
aspects of the Marston theory (Marston, 1913) are reviewed. Then a new method named
"Imperfect Ditch Method of Construction" (Spangler, 1964) are discussed. The imperfect
ditch method utilizes the principles of arch action and arch support to minimize the load
on a buried structure. Figure 3.9 (d) shows the layout of an imperfect ditch conduit.
3.3.2 Assumptions
Loads on a conduit equal the overburden if no relative motion occurs within the
soil or between soil and conduit; however, this is seldom the case. Marston assumed that
sufficient movement occurs to mobilize shearing resistance on sliding planes. After the
movement has been activated, it continues to be effective because of the tendency for
movement, even though the actual finite movements have ceased. The above assumption
was verified by Spangler and Handy (1973) with their in-situ test data. The other as-
sumption that Marston used in his derivations was that cohesion between the backfill ma-
terial and the sides of the ditch was negligible. This assumption yields the maximum prob-
able load on the conduit and offers a safer estimate for our design purpose.
3.3.3 Design of Buried Conduits
In the development of load on an underground structure, arch action is considered
to be the resultant of lateral thrust and vertical shearing forces which are mobilized on
certain vertically oriented planes in the soil overburden. The magnitude of arch support
can be evaluated by means of the Marston Theory. It represents the algebraic difference
between the dead weight of the overburden soil and the earth load to which the structure
is subjected, as indicated by Eq. 3.10. The derivations of the formulas in this section can
be found in Spangler and Handy's book - "Soil Engineer" (1982). They are not shown
in details here.
For ditch conduits, this load formula is derived by considering the forces acting on
a thin horizontal slice of backfill material. The layout of a ditch conduit is shown in Figure
3.11. Equating the upward and downward vertical forces on the horizontal slice, the fol-
lowing equation is obtained:
V
V+ dV + 2Kt' dh = V+ yBd dh (Eq. 3.9a)
Bd
in which
V = vertical pressure on the top of the horizontal slice,
dV = vertical press increment,
y= unit weight of backfill,
K = the coefficient of lateral stress (generally, the Rankine's active ratio :
K~= tan 2(45-4/2) is applied),
4' = tan 4' = coefficient of friction between fill material and sides of ditch,
Bd = width of ditch at top of conduit.
This is a linear differential equation, the solution is:
V = Cd Xy x Bd 2 (Eq. 3.9b)
in which
'(
1- e -2KI H/Bd)
Cd K , (Eq. 3.9c)
2KL'
H = the distance from the ground surface to the top of the conduit,
e = base of natural logarithms.
Hence, for the case of a rigid ditch conduit with relatively compressible side fills, the load
on the conduit (We) will be:
Wc = Cd X y x Bd2 (Eq. 3.9d)
Cd is a coefficient which can be calculated from Eq. 3.9c.
The magnitude of arch support is the algebraic difference between the weight of
backfill and the load on the structure. For the case of ditch conduits, this difference is:
As = y x Bd x (H- Cd X Bd) (Eq. 3.10)
in which A. = arch support (support derived form both sides of the ditch). The thin slice
of backfill material in the free body diagram of Figure 3.11 will look like an arch shape
slice when the arching effect is activated in the backfill over a ditch conduit.(Figure 3.10
(a))
The case discussed above is when an underground conduit is stiffer relative to the
soil medium (i.e. a rigid conduit) and the load distribution at the top level of the conduit
would be like that in Figure 3.12. In other words, the stiff conduit takes almost all of the
load from overburden soil. When this occurs, Eq. 3.9d is valid. However, for the case of
a flexible pipe conduit and thoroughly tamped side fills having essentially the same degree
of stiffness as the pipe itself, the value of W, given by Eq. 3.9d might be multiplied by the
ratio BI/Bd, where B, is the outside width of the conduit. The load from the overburden is
distributed uniformly on the conduit and the soil beside it. Therefore, the load on the
flexible pipe would then be:
We = Cdxy xB xBad (Eq. 3.11)
It's emphasized that for Eq. 3.11 to be applicable, the side fills must be compacted suffi-
ciently to have the same resistance to deformation under vertical load as the pipe itself.
The equation can't be used merely because the pipe is a flexible type. In the actual condi-
tion, it's probable that the load on a pipe lies somewhere between the results in Eq. 3.9
and Eq. 3.11, depending upon the relative rigidity of the pipe and the sidefill columns of
soil.
As to projecting conduits, there are two types of them, the positive projecting
conduits and the negative projecting conduit. When a conduit is installed as a positive
projecting conduit, shearing forces also play an important role in the development of arch
action and the resultant load on the structure. In this case the planes along which relative
movements are assumed to occur and on which shearing forces are generated, are the
imaginary vertical places extending upward form the sides of the conduit as indicated in
Figure 3.13 (a) and 3.13 (b). The width factor in the development of an expression for
load is the outside width of the conduit, designated as B-.
The magnitudes and directions of relative movements between the interior prism
ABCD of Figure 3.13 (a) and (b), and the adjacent exterior prisms are influenced by the
settlement of certain elements of the conduit and the adjacent soil. Marston combined
these settlements into an abstract ratio, called the settlement ratio,

a
rsd = (Sm+) -(f + dc) (Eq. 3.12)

Sm
in which
rd = settlement ratio,
s. = compression strain of the side columns of soil of height pB,,
p = projection ratio,
pBC = the vertical distance from the natural ground surface to the top of the
structure,
B. = outside width of the conduit,
s,= settlement of the natural ground surface and adjacent to the conduit,
sf = settlement of the conduit into its foundation,
dc= shortening of the vertical height of the conduit.
Marston also defined a critical plane, which is the horizontal plane through the top
of the conduit when the fill is level with its top, i.e. when H = 0. During and after con-
struction of the embankment, this plane settles downward. If it settles more than the top
of the pipe, as illustrated in Figure 3.13 (a), rd is positive; the shearing forces on the exte-
rior prisms move downward with respect to the interior prism. The shearing forces on the
interior prism are directed downward, and the resultant load on the structure is greater
than the weight of the prism of soil directly above it. This case is called "projection con-
dition" and the arching effect increases the load on the conduit. If the critical plane settles
less than the top of the conduit, like the one shown in Figure 3.13 (b), red is negative; the
interior prism moves downward with respect to the exterior prisms. The shearing forces
on the interior prism are directed upward, and the resultant load on the structure is less
than the weight of the soil above the structure. This case is called the "ditch condition"
and the arching effect decreases the load on the conduit.
Using the aforementioned parameters, if the shear stresses at the sides of the inte-
rior prism are developed to the top of the embankment, which is called the "complete
condition", the load on the positive projecting conduits derived by Marston is:
We = Cc x Y x Bc 2 (Eq. 3.13a)
in which
C = 2 , (Eq. 3.13b)
+2Kii
p = tan ) = coefficient of friction of fill material.
The plus signs are used for the projection condition and the minus signs are used for the
ditch condition.
The two different results in the analyses of positive projecting conduits are caused
by the stiffness of the buried conduit. If the conduit is rigid relative to the refilled soil, the
projection condition exists like Figure 3.13 (a). If the conduit is flexible relative to the
refill soil, the ditch condition exists like Figure 3.13 (b). Figure 3.14 presents the differ-
ence of the loads on conduits which have different stiffness while under the same situation.
The load on the concrete pipe (rigid pipe) is consistently greater by about 50% than that
on a parallel corrugated steel pipe of approximately the same diameter, i.e. the concrete
pipe can take more load than the corrugated steel pipe.
In order to reduce the load on the conduit under the projection condition. Mar-
ston proposed the idea of the imperfect ditch conduits in 1920. Spangler provided the
analysis of loads on imperfect ditch conduits later in 1950. In the imperfect ditch conduit
construction procedure, illustrated in Figure 3.9 (d), the conduit is first installed as a posi-
tive projecting conduit. Then the soil backfill at the sides and over the conduit is com-
pacted up to some specified elevation above its top. Next, a trench of the same width as
the outside horizontal dimension of the pipe is excavated down to the structure and re-
filled with very loose, compressible material, e.g. loosened soil, straw, or hay. The pur-
pose of this method is to insure that the interior prism of soil will settle more than the ex-
terior prisms, thereby generating friction forces which are directed upward on the sides of
the interior prisms. The resultant load on the conduit is then reduced.
The load formula for an imperfect ditch conduit is :
We = Cn x y x B e (Eq. 3.14)
where C. is a load coefficient which is a function of the ratio of the height of fill to the
width of ditch, H/B,, the projection ratio p', and the settlement ratio rad. Spangler and
Handy provided several sets of C, diagrams with different parameter values (Spangler &
Handy, 1973 & 1982). The settlement ratio, rad, is always a negative quantity in the imper-
fect ditch conduit case, which means the arching effect will always transmit the load at the
top of conduit to the side soil media.(The direction of shear forces at the sides of the inte-
rior prism is always upward.)
The negative projecting conduit has the same function as the imperfect ditch con-
duit. The analysis of loads on negative projecting conduits follows the same procedures as
that for imperfect ditch conduits, but uses a different width factor, Bd, instead of the width
of the imperfect ditch, B,. Bd is the width of the shallow ditch in which the pipe is in-
stalled (see the layout at Figure 3.9 (c)). The same load coefficient diagrams are applica-
ble to both case.
3.3.4 Discussion of the Coefficient of Lateral Stress (K)
In the original formulation, Marston suggested K to be the active Rankine ratio
(K,) in his design of underground conduits. However, the other researchers, such as
Krynine (1945), Ladanyi and Hoyaux, and Handy (1985) found that since K. was derived
from the assumption that the horizontal and vertical stresses are principal stresses, using
K. is only valid when there is no friction (shear stress) between the fill material and the
sides of the ditch. However, Figures 3.10, 3.11, and 3.13, present that there is friction
existing at the sides of the ditch. Soil arches formed in the ditch are caused by this fric-
tion. Therefore, the active Rankine ratio (K.) cannot be used.
If the vertical sides of the ditch are where the friction of the soil is mobilized, then
a new value of K can be determined from Figure 3.15 (Iglesia et. al., 1990). In this figure,
the distance from the point of origin in the r-o coordinate to the center of the Mohr circle
is OC = 1/2-.(h+ov). If the stress at the vertical sides of the ditch lie tangent to the failure
envelope, the radius of the Mohr circle is R = OC-sin4. Since the radius R can also be ex-
(ov - Gh)
pressed as ( s , therefore:
2sine
(ov - oh) -21 (oh +av)sin

(av-a (Eq. 3.15)
2sin 2
Eq. 3.15 can be rearranged to:
oh = KK-'v (Eq. 3.16)

in which
1- sin' ( 2
KK= sin(Eq. cos 24 3.17)
1+sin2 1+sin
3.4 Silo Theory
The formula for computing the pressure/force acting at the bottom of a silo was
first developed by Janssen in 1895 (Jakobson, 1958). Using the layout in Figure 3.16
(Evans, 1983), a silo full of granular material with diameter B and height H, one can de-
termine F, the vertical force on the base. Consider the forces acting on the horizontal dif-
ferential element of height d~,diameter B, and at depth h. Vertical forces on the element
are the downward directed force on the top (V), the upward force on the bottom (V+dV),
and the element's self weight (W = y7B 2dh/4). The lateral stress, ah = 4KV/hB 2, is sym-
metric about the centerline, producing no net force on the element (where K is the coeffi-
cient of lateral stress).
The silo theory is based on two assumptions: (1) The coefficient of lateral stress
(K) has the same value at all depths, and (2) the material settles with respect to the side
walls sufficiently to develop shear stresses over the full depth of the silo. Therefore, if the
element is assumed to move downward with respect to the rigid walls of the silo, the up-
ward acting shear stresses developed will be:
4KVtan #'
S= 4KVta (Eq. 3.18)
rB2
in which tanV' = the coefficient of friction between the granular material and the silo's
walls. These shearing stresses contribute to an upward acting vertical force on the ele-
ment which is equal to rxdhxitB. The vertical equilibrium is therefore:
4KV tan@'dh ynB2dh

V+dV+ 4KVt =V+ dh (Eq. 3.19)
B 4
Solving this linear differential equation:
V =1 tB3 (1-e -4K t w () (Eq. 3.20)

16Ktan '
The force acting on the base of the silo (F) is obtained by substituting H (height of the
silo) for h in the above equation.
The arching effect decreases the force exerted on the base of a silo since all the
shear stresses between the side wall of the silo and the granular material in the silo are
upward. It was also assumed that the vertical normal pressure was uniformly distributed.
But in fact, according to the experiments performed by later researchers, this assumption
was incorrect. Hence, Jakobson proposed a new method to find out silo pressure. He
assumed that every point of the contained mass settles vertically when the load increases,
which means that the ratio between the horizontal normal pressure and the vertical pres-
sure at every point is equal to the coefficient of earth pressure at rest (Jakobson 1958).
Under this condition, uniform vertical normal pressure distribution is unnecessary. He
used the equilibrium equations in a cylindrical coordinate to find the normal force at the
base of a silo:
&rie
-&+ --1 aae + -atez + 2rre
- 0 (Eq. 3.21a)
dr r M 8z r
8trz 1&8 teZ &z trz
+--+-+- = (Eq. 3.21b)
ar r &z r
dOr 1 &fe &Srz or -

-+ -- +-+-=
8r r8 az r
0e
(Eq. 3.21c)
,in the case here:
or = = Oh = K.oz (Eq. 3.21d)
rre = eOz = 0 (Eq. 3.21e)
Trz = T (Eq. 3.21f)

where
r = the vertical and (horizontal) shear stress,
r = the distance from the center of the silo,
oz = vertical normal pressure at the depth z,
K = the coefficient of lateral stress,
y = unit weight of the contained mass.
Substitute the parameters in Eq. 3.21a to Eq. 3.21. Then, Eq. 3.21 can be rearranged as
follows:
K+. +-=y (Eq. 3.22a)

8r z r
K -- +-= 0 (Eq. 3.22b)

ar az
,with the boundary conditions:
C= oz = 0, for z = 0, (Eq. 3.22c)
I= tan' xKxoz, for r = R (the radius of silo) (Eq. 3.22d)

The solution of these equations is very complicated. Jakobson found the solution
by the use of expansion series. Some values of the vertical normal pressure close to the
wall (r=R) and in the center (r=0) were calculated. The calculated vertical normal pres-
sure at r=R agreed with the result from Janssen's method (a, = V/xR2), which assumed
the vertical normal pressure was uniformly distributed. This suggested that Janssen's
method still could be utilized under some situations, although his assumption of normal
stress distribution was inappropriate.
Jakobson (1958) also discussed the ratio between the horizontal normal pressure
and the vertical pressure (K). He believed that this ratio must amount to a certain mini-
mum value which is greater than the ordinary coefficient of active pressure (Ka), even
when the silo walls yield a lot.
3.5 Ground Arch Approach and Ground Dome Approach
3.5.1 Background
Some general principles about underground conduits were outlined by Marston
and his associates (Spangler, 1964). However, there was little information pertaining to
predicting the collapse loading for a soil-structure system involving a flexible tube, arch,
or dome. Some of the important questions left unanswered were as follows: If the struc-
ture yields suddenly, can an arch "immediately" form in the surrounding earth, and also
what strength and stiffness is necessary in the surrounding soil to ensure that a thin arch
will fail by compressive yielding rather than by buckling or bending? In order to find the
answers, Whitman, et. al. (1962 & 1963) made observations from small-scale tests on thin
metal domes buried within a coarse dry sand (static loading). They found that the level of
pressure required to cause failure of the buried dome was several times that required for
failure of an unburied dome. They postulated that a sand dome had developed above the
structure, assuming some of the load which had previously acted upon the structure. This
explained the reason that higher pressure was needed to cause the failure of a buried
dome.
3.5.2 Ground Arch/Dome Approach
After Whitman et. al. (1962 & 1963), several researchers proposed a similar
"structural" hypothesis to explain the load redistribution around a yielding structure. In
the planar condition, they suggest that when the adjacent soil deforms or yields, an arch or
ring will form within the soil. In the spatial condition, a dome or sphere will develop in
the soil. Once these "ground-structures" are developed, they will redistribute loads away
from the actual buried structure and only part of the original overburden pressure will be
transmitted to the structure itself.

Figure 3.17 shows some examples. These soil arches and domes were proposed
by Nielson in 1966. The differential soil arch in Figure 3.17 (a) was used to determine
arching over a buried conduit. The arch was assumed to be circular, with supports located
on the surfaces of maximum shear stress (these surfaces are determined by the elasticity
theory). Then the problem was solved using numerical procedures and a computer solu-
tion (code) was presented. Later on, Nielson extended this approach to some non-circular
structures. A rectangular underground structure is shown in Figure 3.17 (b). He also

applied the approach to a three-dimensional structure(Figure 3.17 (c)). In the three-
dimensional case, a differential soil shell was formed. This was accomplished by putting a
soil dome above the underground structure to assume part of the overburden pressure at
the top of the structure. However, no attempt was made to formulate the solutions for
the rectangular case and the three-dimensional case (Evans, 1983).
As to the increased load-carrying capacity of the buried structure, Luscher and
Hoeg (1964 & 1965) attribute it to three types of action by soil around the buried struc-
ture: pressure redistribution, deformation restraint, and arching. They proposed a study to
analyze the aforementioned interactions between a buried cylindrical tube and the sur-
rounding soil under large applied loads. These three types of action are briefly explained
as follows:
(1) The restrain against tube deformations in the second mode (that is, counteract-
ing the deformation from the originally circular shape into a horizontal ellipse, see
Figure 3.18) by mobilization of lateral passive earth pressures is called "pressure
redistribution" (Luscher & H6eg, 1965). In the other words, this restrainment
causes stresses to redistribute to a more uniform configuration as the crown de-
flects vertically and the springlines horizontally.
(2) The action against deformations in the third and higher modes (see Figure
3.18) enhances the resistance of the tube against buckling failure by forcing it to
buckle in higher modes rather than in an unsupported situation: this action is called
"deformation restrain" (Luscher & H6eg, 1965). The buckling resistance of the
tube is increased dramatically here, i.e. the load-carrying ability of the tube is in-
creased greatly at this process.
(3) The reaction of the surrounding soil to tube deformations in mode one (pure
compression) or mode zero (rigid body motion) is called "arching". Redistribution
of pressure away from or onto the tube, depending on the relative compliances of
tube and soil surrounding, is defined as active or passive arching, respectively
(Luscher & Hoeg, 1965).
The distinction between these three actions was somewhat artificial and arbitrary.
Because the effects operate simultaneously, they influence each other. In Luscher and
Hoeg's experiments on buried flexible tubes, they concluded that a sand ring, as shown in
Figure 3.19 formed around the tube when deflection occurred. They believed that the
vertical-sliding-surface concept (Figure 3.5 and Figure 3.6) from Terzaghi (1943) should
be replaced by the concept of thrust-ring action, i.e. structural arches or domes forming in
the soil. Only dry, cohesionless soils were specifically considered in their research. How-
ever, Luscher and Hoeg believed their conclusions apply in a general way to any soil.
There are also numerous field data supporting the contention that buried structures
have a much higher load-carrying capacity. These data were presented by Davis and
Bacher (1968) in their California's Culvert Research Program. They gave magnitudes of
actual loads experienced and ultimate capacities of culverts installed using several different
backfills and backfilling techniques as well as load capacities of unburied culverts.
Another structural analogy for arching was also proposed by Getzler, et. al. (1968)
in experiments which they conducted with structures having various roof shapes. Their
experimental setup is shown in Figure 3.20. The analytical model they utilized is shown in
Figure 3.21. They proposed a soil arch on the top of the buried structure. This arch sits
on the soil abutments, which are the soil bodies at the side of the structure. However, in
their study, no theoretical formulation was presented. They reached a conclusion in their
experiments about the effect of different roof shapes of buried structures. They summa-
rized that buried structures with triangular or arch shape peak roofs, experienced greater
load reduction from arching than structures with flat roofs. Similar experimental results
from Whitman et. al. were found in 1962. In Whitman's tests a flat roof experienced ap-
proximately 50% higher total load than a hemispherical roof at the same depth of soil
cover. Luscher and Hoeg (1964) also discussed this problem in their study. They be-
lieved the roof shape was an important factor to be considered when the stress acting on a
buried structure was analyzed. The vertical-sliding-surface analysis does not take into ac-
count the geometric configuration of a buried structure but the trust ring method (Figure
3.19) takes into account the shape of the buried structure.
The above soil structural approaches provided explanations for the increase of the
load-carrying capacity of the buried structures and also showed the interaction between
the structure and soil body. Nevertheless, these approaches are rarely used in actual ap-
plication. Selig (1975) released a paper to summarize the basic concepts of soil-structure
interaction and the nature of the stresses and deflections associated with buried structures.
He used the large corrugated-metal buried structures, which are commonly referred to as
"flexible' structures, to verify different approaches of the "soil-structural" analysis. He
pointed out that the aforementioned approaches have never been fully developed. In order
to obtain the results, engineers have to make a lot of assumptions such as the shape of the
arch or dome, locations of the supports, conditions at the supports ,and interference be-
havior between elements. However, with different assumptions, the results may vary a lot.
These assumptions give the risk and inconvenience to the engineers when they use these
approaches.
TM
Sand
31cm
trap
door L
17.3cmi
Figure 3.1 Terzaghi's Experimental Set-up (Terzaghi, 1936)
15.0
12.5
o
o 10.0
S9.6
x4~ o0
Dense
6.0 sand
5.0
5.0 10.0 6 15.0
Normalized trap door displacement(S x 10o)%
'Displacement ratio'
Figure 3.2 Terzaghi's Experimental Results: Vertical Force on Trap Door vs. Displace-
ment of Trap Door (Terzaghi, 1936; Revised by Evans, 1984)
0.02 0.04
0.01 0.03 0.05 0.01 0.03 0.05
dv in kg per sq. dh in kg per sq.
Figure 3.3 Terzaghi's Experimental Results: Vertical and Horizontal Stresses in Soil Body
vs. Depth (Terzaghi, 1936; Revised by Evans, 1984)
ZB
0
1.0 2.0 3.0
Coeff. of lat. earth press. (K)
Figure 3.4 Terzaghi's Experimental Results: Coefficient of Lateral Earth Pressure (K) vs.
Depth (Terzaghi, 1936; Revised by Evans, 1984)
I
!/
So/id bcs5e
Figure 3.5 Yielding in Soil Caused by Downward Movement of a Long Narrow Section
(ab) at the Base; Curve ac & bd: Actual Sliding Surfaces, Line ae & bf: As-
sumed Sliding Surfaces (Terzaghi, 1943)
-ZB --I
7,;
Figure 3.6 Free Body Diagram for a Slice of Soil in the Yielding Zone (Terzaghi, 1943)
I
0
I 0_._.--
(a) (b)
Figure 3.7 (a) Flow of Soil Toward Shallow Tunnel When Yielding Happened in the Soil
Body; Curve eb1 : Actual Sliding Surface, Line elb1 : Assumed Sliding Surface,
(b) Vertical Stress Profile in Soil Located above the Tunnel (Terzaghi, 1943)
0
..
DP
I~2
D1
i I
(a) (b)
Figure 3.8 (a) Yielding Zone in Soil When Tunnel Located at Great Depth, (b) Vertical
Stress Profile in Soil Located above the Tunnel (Terzaghi, 1943)
Natural ground surface \ Top of embankment
-rpl~nmrrnmnmtm~iiilr1nrr.Inlrrl!lnl~lH71
t1
c
E
,
u
Natural
LOA r~.-.- A'.

h ground\
(a) (b)
/ Top of embankment
L--.J
.........................................
m ppr! .. , "irrr. 111t,1
17r7 r i /HI tr~/ar/tfr /lr t.'" '"l
fll r ". .. " "
Excavate and refill

Natural ~rnund with loose soil
Compacted Lt _ ]J Compacted
(d)
Figure 3.9 Various Classes of Conduit Installations: (a) Ditch Conduit, (b) Positive Pro-
jecting Conduit, (c) Negative Projective Conduit, and (d) Imperfect Ditch
Conduit (Spangler &Handy, 1973)
I-
4)
.- 4
4)4
C
0
h
oN
Ic
0
cu
O
a)
I i
ob
OC r
pa
o
C= ';3
.-
tr <
Us
80
"B
Natural ground
dhl
Figure 3.11 Free Body Diagram for Ditch Conduit (Spangler & Handy, 1973)
Figure 3.12 Load Distribution at Level of Top of Pipe (when sidefill soil columns are more
flexible than the pipe, from Spangler & Handy, 1973)
-o
Eo
E
.0
E
0
0.
0P
E
C
E
0,
0.
0
Marston load
IOO 140
1118 mm (44 in.) o.d. concrete pipe
120
Weight of soil prisms oomm
above tops of pipes
100
"'400 1082 mm (43 in.) o.d. corrugated steel pipe

i i---i i i i
o
I E
0
r I '
f' E .
: ~i
I I [0
V
1 2 3 5 7 12 17 21
Time (yr)
Figure 3.14 Comparison of Measured Loads on Rigid Pipe and Flexible Pipe (Spangler &
Handy, 1973)
a
Oh = Kx
1 - 0" wY 1 +sin22
1KK
+ sin22
Figure 3.15 The Coefficient of Lateral Stress (K) for the Design of Underground Conduits
(Iglesia et. al., 1990)
dh
dh
Figure 3.16 Free Body Diagram for Silo Theory (Origin form Janssen (1895), the picture
shown here is the revised plot from Evans (1984).)
Pt
Pt 1
(a) Free body for a buried conduit
ture
(b) Free
rectal
Figure 3.17 Free Body Diagram for Nielson's Arching Analysis (Nielson, 1966)
LI C~
P + dP
S ch support
;oil arch support
(c) Free body for a three-dimensional structure
Figure 3.17 (cont'd) Free Body Diagram for Nielson's Arching Analysis (Nielson,
Original Circulor ShaoDe -
Deformed Shaoe ---
Mode Two Mcde Four
Figure 3.18 Deformation of Tubes: Mode Two and Mode Four (Luscher & Hoeg
Rive
il
S
d
Assume
tube)
Figure 3.19 Soil Arching as A Thrust Rings About the Structure (Luscher & Hoeg, 1964)
F
T
LT
HT
SECTION (PARTIAL)
Figure 3.20 Trap Door Test Instrument with Various Roof Shapes (Getzler et. al., 1968)
SUITACE LOAD
1I I I I I I 1I I II I II I 1
SOIL MIDIUm
Figure 3.21 Soil Arching as An Arch Above the Structure (Getzler et. al., 1968)
Chapter FOUR
Analytical Approaches
4.1 Introduction
In this chapter, several theoretical solutions will be discussed. First are the con-
tinuum approaches which use the elasticity theory and the plasticity theory to study the
stress redistribution in the trap door experiment and arching in underground construction.
Second, several discontinuum methods will be illustrated for comparison with the contin-
uum methods. There is, however, little research dealing with the discontinuum ap-
proaches because of their complexity and because continuum approaches can generally
describe soil behavior quite well.
Numerical analyses utilized to study underground openings are also examined in
this chapter. These analyses are perhaps the most comprehensive methods available for
the analysis of stress redistribution around buried structures. Numerical analyses allow
one to analyze problems with complications such as involved geometries, concentrated
loads, non-homogeneity, and anisotropic behavior. Theoretical solutions will have limita-
tions due to their assumptions which are used to simplify problems. Numerical analyses
are more flexible regarding their assumptions and, therefore, can generate solutions closer
to the real situations. Several numerical analyses used to study arching will be introduced
in this chapter. They include the analyses by the finite different method and finite differ-
ence method. A new numerical simulation using the distinct element method (Sakaguchi
& Ozaki, 1992) is also presented.

4.2 Continuum Approaches Using Elasticity Theory
An underground structure is normally treated with continuum approaches to sim-
plify the analysis of its mechanical behavior. Determination of the stress distribution in
continua is usually based on elasticity theory. In other words, the mechanical behavior of
the geomaterial and of the underground structure are assumed elastic. Some researchers,
however, do not believe solutions based on elasticity theory can describe the real mechani-
cal behavior of ground-structure systems, since most of the deformations caused by the
ground-structure systems are inelastic. This is especially prevalent, when we try to de-
scribe arching phenomena in the ground, since arching effects occur mostly in a ground-
structure system which deforms plastically. This section introduces the elastic solutions.
Finn (1963) presented closed form solutions for the change in vertical stress result-
ing from translation or rotation of a trap door. Figures 4.1 (a) and (b) show the two
boundary conditions in his analysis. Figure 4.1 (a) is the pure translation case and Figure
4.1 (b) is the pure rotation case. In these figures, the vertical displacement (v) and the
vertical stress (Co) on the ground surface are both assumed to be equal to zero. A plane
strain condition is assumed with the soil treated as an elastic medium with unit weight (y)
resting on a rigid horizontal boundary with a trap door located in it. The rigid base is
considered frictionless (r, = 0) initially (Figure 4.1 (a) (b)) but is then treated as frictional
and cohesive later. The trap door has a width equal to 2b. The depth of soil is assumed
to be infinite (h-+oo). The displacement of the trap door is d. Finn restricted his analysis
to problems where displacement of the soil was very small and entirely elastic.
A typical distribution of the change in vertical stress across the base for a down-
ward translating door is shown in Figure 4.2. Infinite tensile stresses develop near the
edges of the trap door, while infinite compressive stresses occur on the base next to the
door. The results obtained by the theory of elasticity from Finn (1963) were checked
against available published experimental and field results. The stress distributions, the ap-
proximate location of the force resultants, and the influence of the various types of dis-
placements predicted by the analysis were in reasonable agreement with the published re-
sults (Finn, 1963). However, one big obstacle to the determination of these values was
the displacement of the trap door, d. These predicted results were only good for very
small d values.
Finn also applied his solution based on elasticity theory to the analysis of retaining
walls. Figure 4.3 shows the layout of a retaining wall which is similar to the set-up in Fig-
ure 4.1 (a). But the trap door is changed (erected) to be a retaining wall. The horizontal
displacement of the retaining wall is equal to the vertical translation movement of the trap
door. The pressure on the wall due to this displacement, however, will only be the pres-
sure on one half of the trap door because the displacement is modified by the effect of re-
moving the stresses on the line corresponding to the free surface of backfill (Figure 4.3).
Based on the aforementioned method, Finn found that the resulting pressure distribution
against a translating wall corresponded to the results of Taylor's method (1948). He sug-
gested a combination of his translation and rotation solutions could be used to predict the
stress distribution of a general retaining wall.

In Finn's study, the depth of the soil was always taken as infinite which imposed
restrictions in adapting the solution to practical problems of finite soil depth. Chelapati
(1964) presented a study using Finn's model but dealt with the stresses in a soil field of
finite depth, h. The soil mass was again assumed to be a homogeneous, elastic, isotropic
medium but subjected to high overburden pressure, i.e. cy = Po.. The geometry and
boundary conditions of Chelapati's analysis is shown in Figure 4.4. Basically, it is the
same as Finn's analysis in Figure 4.1 (a), except in the finite depth (h) and large overbur-
den pressure (Po).
Using the model in Figure 4.4, Chelapati superimposed stresses caused by the
yielding trap door onto those due to surcharge load. The problem of infinite stresses at
the trap door edges still existed. Since he considered granular soils in his study, and
granular soils cannot sustain tension, the stress on the door was assumed to be zero wher-
ever tensile stresses were indicated. The vertical stress distribution across the base for a
downward translating door is shown inFigure 4.5. In the figure, H is the depth of soil and
B is the width of the trap door. Compressive stresses on the base adjacent to the door
were then reduced to produce no net change in total vertical force on the boundary (the
boundary includes the trap door and the base), as shown inFigure 4.5. Chelapati used the
method of series expansion to find the solutions in his study. All results were presented
graphically in terms of h (the depth of soil), b (the trap door width), P (soil self weigh, Po
, plus overburden pressure, yh), E (Young's modulus) , and p (Poisson's Ratio). Chela-
pati concluded from the results that arching for the cases in his study was dependent on
three parameters, b/h, Ph/dE, and p. For practical purposes, however, the effect of Pois-
son's ratio, p, can be neglected over a wide range of the other parameters.
Bjerrum, Frimann Clausen, and Duncan (1972) believed that Chelapati's elastic
solution can be further extended to give approximate values for the change in vertical
pressure (Ap) at the center of a flexible section located within a rigid horizontal boundary.
The layout of the model from Bjerrum, Frimann Clausen, and Duncan is shown in Figure
4.6. The variation of vertical pressure at the center was expressed as:
Ap aox(8/1)xE (Eq. 4.1)
in which,
8 : the deflection of the trap door,
I : half width of the trap door,
8/1 :the deflection ratio,
E: Young's modulus,
a. : a coefficient whose value varies from 0.3 to 1.0 depending on the fac-
tors, h, 1,8, and p,
h: soil depth,
p :Poisson's ratio.
The above equation shows that the pressure changes due to arching increase with E
(Young's modulus). Therefore, the pressure change is greater in dense sand than in loose
sand, and it is greater in sand than in silt or clay. This result agrees with the observations
in real engineering cases. In addition, this formula is restricted to those cases which have
84
small Ap values. Hence, the accuracy of this formula is questionable when large stress re-
ductions are observed in the laboratory and field.
This kind of approach was taken one step further by Burghignoli (1981) when in-
vestigating stress redistribution around rectangular underground openings with flexible
roofs (crowns). In his analysis, he assumed linear elastic behavior both for the soil and for
the roof. Using the arching model from Bjerrum et. al. (1972), Burghignoli treated the
tunnel crown as a flexible part in a rigid base (Figure 4.6). In other words, he incorpo-
rated the elastic property of the flexible section (the roof) within the rigid boundary into
the arching analysis. All of Burghignoli's results were presented in graphical form. He
used the relevant results in the analysis of the interaction between the crown of the tunnel
and the overlying soil. Some appropriate estimates was given in Burghignoli's study to
demonstrate the validity of his approach. Iteration is necessary when using his method to
analyze a tunnel structure, because a deformation (8) at the centerline has to be assumed
first (see Eq. 4.1).
4.3 Continuum Approaches Using Plasticity Theory
Visual observations and pressure measurements show that many of the previously
derived analytical and experimental methods are not correct. For example, soil does not
behave elastically, nor does it shear along vertical planes progressing from trap door
edges. Peck (1975) doubted the applicability of arching analyses which assumed elastic
behavior. He stated "The ground movements associated with construction, particularly in
soft soil, are so large that the soil is likely to be stressed far beyond the limits of elasticity"
(Peck, 1975). Evans (1984) also mentioned the inappropriateness of the elasticity as-
sumption. He believed this assumption would result in predicted thrusts and moments
larger than those actually measured infield and laboratory installations (Evans, 1984).
Since the elasticity theory cannot describe the arching behavior well, Evans (1984)
provided an approach which used plasticity theory to solve the problem of a translating
trap door within a granular soil. Figure 4.7 shows the idealized model for his analysis.
The soil was represented by a frictional granular material which was cohesionless, iso-
tropic, and homogeneous. A layer of this material with finite thickness (H) and unlimited
lateral extent rested on a base containing a trap door. This door which might be raised or
lowered, had width B and an infinite length perpendicular to the plane of the figure, i.e.
plane strain conditions were assumed. His model is similar to those using elasticity theory
where the traditional stress-strain equations from elasticity theory govern the soil's behav-
ior in the elastic zones. However, here the failure criteria and plastic flow rules define the
behavior within the plastic zones. Evans utilized the Coulomb failure criterion and the as-
sociate flow rule in his development of plastic solutions for the trap door problem.
In the plastic solutions of Evans, two major factors govern the deformation of a
soil body with a trap door at the bottom. The first factor is the direction of the trap door
movement. If the trap door moves downward away from the soil body, then this is the
active case. If the trap door moves upward into the soil body, then this is the passive case.
The other factor is the variable called the "angle of dilation (v)". The angle of dilation is
the angle of the plastic potential with respect to the horizontal. The "plastic potential" is a
curve defined as being perpendicular to all plastic strain increment vectors. Hence, the
86
flow rule forms vectors in the 86,Pfl -6sic space as shown in Figure 4.8. "6" in Fig-
ure 4.8 represents the changes of these two strains. v also defines the direction of the
plastic strain increment vector and is therefore a convenient variable to use for expressing
a plastic flow rule. When v is 0, no volume change occurs, while for v > 0, the material
will dilate or expand.
By using the plasticity theory, Evans discovered different theoretical results. In
active arching, when the angle of dilation (v) is larger than zero, a prism of soil above the
trap door will move downward and the shape of the soil prism is triangular, like Figure 4.9
(a). However, when v is equal to zero in the case of active arching, the shape of the soil
prism is rectangular (Figure 4.9 (b)), like the result from Terzaghi (1943). The similar
condition happens in passive arching. When v is larger than zero, a trapezoidal prism of
soil moves upward (Figure 4.10 (a)). When v is equal to zero, a rectangular soil prism is
used to analyze the arching problem. Evans expanded his theoretical solutions to three-
dimensional cases after these two-dimensional studies. He considered different geometries
of the trap door in his three-dimensional cases. The shapes of the soil prism used in his
three-dimensional solutions are shown in Figure 4.11 to Figure 4.14. Figure 4.11 and Fig-
ure 4.12 present the cones of soil above a circular trap door. The active case is a cone
(Figure 4.11) and the passive case is a truncated cone (Figure 4.12). Figure 4.13 and Fig-
ure 4.14 show the soil prisms above a rectangular trap door.
Evans assumed a perfect plastic soil model in his arching analysis. However, the
real soil is not a perfect plastic material. Therefore, his plastic solutions for arching are
not very rigorous but these solutions do provide insight into what is actually occurring
87
within the soil when a trap door is lowered or raised. Most of the theories derived by
Evans assumed plane strain behavior (like Figure 4.9 and Figure 4.10). These derived
theories were compared with the experimental results in Evans' study. The comparison
proved that plasticity theory could give a better description of the ground deformation be-
havior and the arching effect in granular soil. More details are discussed in Chapter Six.
As to the three-dimensional case, since preceding literature contains little information on
the three-dimensional behavior of particulate matter, it is difficult to evaluate the appro-
priateness of Evans' extensions for the circular and rectangular trap doors.
4.4 Discontinuum Approaches
The ideal approach to arching in granular soil is to solve the equilibrium of individ-
ual soil particles. This kind of method is called the discontinuum approach. The discon-
tinuum approach is not practically applicable in the study of arching in granular soils, be-
cause of the vast amount of soil particles. However, Trollope (1957 & 1963) derived a
solution for arching beneath a triangular embankment on a flexible base by this kind of ap-
proach. He believed that the stress distribution induced by body-force within granular
masses could be evaluated from an analysis of the static equilibrium of a systematically
packed system of mono-sized, smooth, rigid spheres. Trollope postulated a "systematic
arching theory" in two dimension. In this theory, soil was represented by single sized
discs in a regular packing. Figure 4.15 shows the packing. Each particle was acted upon
by six normal forces from adjacent particles (Figure 4.15). The whole soil body was then
88
solved as a static equilibrium system with the assumptions that horizontal forces were zero
and no tensile forces existed.
The results from Trollope's solution showed close agreement with the measured
values in his experiment of the laboratory scale and full size embankments. Butterfield
(1968) utilized Trollope's concept inhis studies on the stresses developed within silos. He
produced results which were similar to those from the silo theory (Section 3.4), but he
failed to give enough information as to how his model was set up.
Maeda et. al. (1995) utilized the discontinuum approach to study the deformation
and failure mechanism in granular materials. They used a microstructure composed of
granular particles aligned as an ellipse as a basic unit (Figure 4.16). When they modeled
the elliptic microstructure, the shape, size, and number of constituent particles were not
taken into account; only the shape of the elliptic structure was considered as an essential
parameter controlling the mechanical properties of the structure.
The mechanical properties of this elliptic structure were developed in two dimen-
sions. Interparticle contact forces inthe elliptic structure were found by the static equilib-
rium in each microstructure. As shown in Figure 4.16, the contact force (f) can be de-
composed to normal component (Q)and tangential component (f). The relation between
these two force components is:
ft=fn x tan40 (Eq. 4.2)
in which 4, is the mobilized interparticle friction angle.
Each elliptic microstructure was subjected to the external contact forces generated
at interface with adjacent elliptic structures. The external contact forces between elliptic
89
microstructures were calculated. Then these contact forces were "homogenized" to the
"equivalent stresses" as shown in Figure 4.17. Maeda et. al. used this idealization
(homogenized) to develop the macroscopic mechanical models of granular materials.
With this elliptic structure model, Maeda et. al. found that the stability of the soil body
could be presented as a function of the shape of the elliptic structure and the stress condi-
tion surrounding the structure. They also believed that the characteristic mechanical prop-
erties of a granular material, such as deformation anisotropy and dilatancy, could be ex-
plained by the elliptic structure, where the rotation of the constituent particles played an
important role. However, there were not many analysis results available at the present
time. The applicability of their model to different types of granular materials is still un-
sure.
4.5 Numerical Methods
Due to the simplified assumptions, the aforementioned theoretical approaches are
usually limited in the application of practical engineering. Compared to the theoretical
approaches, numerical methods are more flexible, and may be the most comprehensive
methods available for the analysis of stress redistribution when dealing with arching prob-
lems. Several kinds of numerical methods were used by engineers and researchers, includ-
ing: Finite Difference Method (FDM), Finite Element Method (FEM), Boundary Element
Method (BEM), and Distinct Element Method (DEM; also called "Discrete Element
Method"). These methods allow for analysis of problems with difficult boundary condi-
tions or complicate material properties, such as irregular geometries, non-homogeneity,
and anisotropic behavior.
Getzler, et. al. (1970) performed a finite difference analysis of plane strain trap
door problem. The soil was assumed to be linear elastic and Chelapati's theoretical solu-
tion (Section 4.2) was used as the formulation in their analyses. Figure 4.18 presents their
computational model. The roof of the underground structure is a trap door located in a
rigid layer. Figure 4.19 shows the superposition of two different modes in their model.
Mode (a) is no settlement of the trap door relative to the medium, in which the external
surcharge and overburden stress are taken into account. Mode (b) has the differential
settling of the structure, with no external surcharge or overburden stress. From the nu-
merical analysis results, Getzler et. al. found the principal compressive stress trajectories
produced by the differential settling indicate that a soil arch was formed in the overlying
soil, abutting on both sides of the structure and transferring part of the load to those
zones. These discoveries agree with the theory which interprets the arching effect as a
"structure-like" action of the overlying soil, for example, an arch in plane strain case
(Whitman, et. al., 1962 & 1963; Nielson, 1966) and a dome in three dimensional case
(Nielson, 1966).
Ranken and Ghaboussi (1975) used an axisymmetric finite element program to
simulate an advancing tunnel in isotropic, homogeneous, and continuous soil. Again, the
soil was treated as a linear elastic material. Several cases were run in their analysis using
combinations of the following conditions: (1) lined and unlined tunnel, and (2) linear elas-
tic soil, elasto-plastic soil with strength independent of mean stress and angle of friction,
and elasto-plastic soil with strength dependent on the aforementioned parameters. They
obtained reasonable data from their finite element program and suggested various values
for different tunnel designs in their paper.
Stone (1988) performed non-linear finite element analyses of the trap door tests to
help explain his experimental outcomes. He reproduced the characteristics of his physical
model tests with numerical analyses which resulted in a correct prediction of an initial soil
localization propagating from the edge of the trap door into the overlying soil. The strain
distributions simulated numerically were seen to form an arch at small door displacements.
However, the discrete nature of the formation of the soil localizations observed in his
physical model did not show in the numerical analysis result.
Koutsabeloulis and Griffiths (1989) considered the soil as a perfectly plastic mate-
rial and ran a finite element analysis of the trap door problem. The Coulomb failure crite-
rion was used with the non-associate flow rule. The soil body was discretized to fifteen-
node iso-parametric triangular elements, as shown in Figure 4.20. Figure 4.21 presents
the predicting results of their analyses for active and passive arching, in which H is the
depth of overburden soil and D is the width of the trap door. In the active arching case,
the load on the trap door drop down as the trap door is lowered. However, after the load
drops down, instead of an increase, the load just stays at a constant value. This phenome-
non is different from most of the trap door test results. As to the passive case, the curves
show a similar tendency as physical test results. The parameter H/D shows the same ef-
fect in these numerical results as the former testing results. As the value of H/D increases,
the arching effect is more apparent (i.e. the load on the trap door decreases more in active
arching and increases more in passive arching).
A new numerical method called "Distinct Element Method" (or "Discrete Element
Method") considers granular material as a constitution of many individual particles. The
distinct element method can give a better description of the discrete characteristic of a
granular material due to its special capability of modeling each particle as a separate entity.
Sakaguchi and Ozaki (1992) used this method to analyze the formation of arches plugging
the flow of granular materials. Their analyses were applied when the granular particles
discharging out of a silo due to the gravity. The granular particle was modeled as an elas-
tic disc which size is 10 mm in diameter. A concept called "rolling friction" which allowed
the effective development of geometrical interference among particles, was introduced in
their analyses. Figure 4.22 presents their simulation results: the upper plots are the trans-
forming configurations of particles in a silo, and the lower plots are the pictures showing
the velocity vector fields. The arches were formed in the final stage (1.28 sec) and
plugged the particle flow. Sakaguchi and Ozaki also ran several tests to compare their
numerical results. The comparison showed that the prediction from the distinct element
method was in good agreement with the experimental measures not only in the flow pat-
tern but also in the formation of arches.
Numerical methods are recently applied widely in engineering study and practice,
because they are flexible and because confronting problems are more and more compli-
cated. However, there are still some shortcomings in numerical methods. One is that
these analyses require appropriate selection of input parameter values, and sometimes, a
constitutive model (e.g. Cam Clay Model) to obtain good results. The second is that these
methods are usually time-consuming and expensive, especially when one uses trial and er-
ror method to adjust the input data. Hence, a simple and quick method is necessary for
rough estimation and preliminary design in complicated engineering problems. This kind
of method is called empirical method which is discussed in next chapter.

SURFACE OF V=O, 0o=0
SOIL MASS j__l_ __
y
/ l illy l/// -/
///11 /1/o lI// 71
y hm --/o
77////777-77
WT. OF SOIL PER CUBIC FOOT = Y
SMOOTH
. I-I.-LI.IM2 VAI.=
(a) Translational Yielding Base
v-- O
9177T7177771IP7117
yO
I I
RIGID Id I d gx
BAS
I7 7YIELDI777G
TTRYIELDING
STRIP
(b)Rotational Yielding Base
Figure 4.1 Boundary Conditions for Soil Mass with Yielding Base (Finn, 1963)
Compression
Tension
I I
Trap door
Figure 4.2 Typical Distribution of Change in Vertical Stress for Downward Translating
Trap Door (Finn, 1963; Revised by Evan, 1984)
d- k FREE SURFACE
S OF BACKFILL -
i
y
1T//11111/7/1///1111//I/I////111
-- CORRESPONDS TO xio
IN FIGURE 4.1 (a)
*--TRANSLATING WALL
(CORRESPONDS TO HALF
WIDTH OF TRAP DOOR)
7"//1TIiiT/1111111111 i111 1
DEFORMATION OF THIS --
LINE ASSUMED NEGLIGABLE
COMPARED TO d
Ix
Figure 4.3 Translating Retaining Wall (Finn, 1963)
=%
C
(7
-b--cb-c
r" 4e
Figure 4.4 Boundary Condition for Chelapati's Analysis of Arching in Granular Materia
(Chelapati, 1964)
Ground Surface
"ess
:ile
ns i
Figure 4.5 Chelapati's Technique for Elimination of Tensile Stresses using an Elasi
lution for Arching (Chelapati, 1964; Revised by Evans, 1984)
Sodl
21
Ap %w , E ( p small compared with p, a constant)
Figure 4.6 Arching Model in the study of Bjerrum et. al. (Bjerrum et. al., 197
Gr.nalar soil
*A, pl(c=o) . . H
LUnyielding B -~ Trap door

surface
Figure 4.7 Idealized Model for Trap Door Problems (Evans, 1984)
Uplastic
6
shear Positive 6ol= Contraction
Dilation
(v> 0)
'
Constant
volume K
f
.
V=O0
0 0<V<o
( 6 plas tic
shea .r
(-6eplastic)
VUL
lastic potential
6,plastic
vol
Figure 4.8 Plastic Flow Rule (Evans, 1984)

i
cos vsinO
=n 1-sinvsin7
dh
N=j n cos v
T= dh
cos V
A=Area of element
I B J
i' -;
A. Free Body Diagram for V>Oo
~ -~ -- - --
~ I
. . . .
T=:nsino V
4
N=cndh
yBdh dh
T T
T= Tdh
V+dV
TRAP tFDOOR
B. Free Body Diagram for v=0 0
Figure 4.9 Free Body Diagrams for Active Arching in the Two-Dimensional Case
(Evans, 1984)
100
dh
n cosy
T=
dh
----
co san
COS'
n 1-si
A=Area
I B I
A. Free Body Diagram for v>0 0
N=Crd:
T =Tcdh
d=5ns
I B _/
0
3. Free Body Diagram for v=0
Figure 4.10 Free Body Diagrams for Passive Arching in the Two-Dimensional Case
(Evans, 1984)
101
-!
I 1
Figure 4.11 Soil Cone for Active Arching Above a Circular Trap Door (Evans, 1984)
102
H
Figure 4.12 Soil Cone for Passive Arching Above a Circular Trap Door (Evans, 1984)
103
i 1
00
CO
0\
L.
I-
0
L..
L.
104
00
CU
o
0
CL
co
0
c
0
U,
cL
0U,
105
|
Single-sized two-dimensional particles
Figure 4.15 Model for Systematic Arching Theory (Trollope, 1957; Revised by Evans,
1984)
106
~
rticle
fI
Plane
nt
Figure 4.16 Geometrical Definition of Elliptic Structure (Maeda et. al., 1995)
Principal Stress
External Contact Forces
Aur
Homogenization
Fo
Figure 4.17 Homogenization of External Contact Forces with Stress (Maeda et. al., 1995)
107
4b
Figure 4.18 Computational Model of the Finite Difference Analysis (Getzler, et. al., 1970)
D., D.,
Mode (a) Mode b)T

Figure 4.19 Superposition of Two Different Modes (Getzler, et. al., 1970)
108
E , 130 000 kN/m 2
3
y - 20 kN/m
(a) L/B = 5
(b) L/B = 10
Figure 4.20 Finite Element Discretization of the Trap Door Problem

(Koutsabeloulis and Griffiths, 1989)
109
Active mode
Plane strain
=-20T
to. W&l
0-75-
I
0.50-
025 S .4
-.--
5
0i
5I
0 1 2 3 4 5
Displacement:m x 10- "
1 PPasrve mnod
Plane strain HID 5
0 = 330
3. p=00
-- "
- - - - - - - - - -
Ko cb1 ------ ------- - 3
0 . .-
:.....,
-..... :
11
oJ
0 1 2 3 4 6
Displacement:m x 10- 3
Figure 4.21 Numerical Analysis Results (Koutsabeloulis and Griffiths, 1989)
110
-. U
aL
/?'O
SI, I
0,
_ _ _ _ 75
03
1s
Il)
-J
111
Chapter FIVE
Empirical Methods
5.1 General
For the design of underground openings, there are some simplified methods which
basically evolve from previous construction experience. Most of these empirical and semi-
empirical methods indicate a value of the pressure to be applied to the support system of
the underground opening. The shape of openings, after a failure on the roof has taken
place, in many instances resembles an arch. This chapter will illustrate several empirically
based methods and explore the concept of arching effect embedded in them. Some have
been obtained from very simple static analyses and others were entirely founded on the
experience of the person proposing it.
5.2 Classification of Empirical Methods
The empirical methods generally use the assumption that a zone of loosening exists
above the underground structure. The loosening zone may have different shapes depend-
ing on the assumptions and derivations in a method; for instance, inFigure 5.1, the loosen-
ing zone is bounded by a parabola. In this method, the force acting on the structure is as-
sumed to be the weight of the material within the loosening zone plus surcharges.
In general, there are two approaches when determining the loads on the supports
of underground openings. One approach is to take the effect of overburden depth into
account, and the other is to disregard the effect. Consideration of the overburden depth
112
m
depends on whether the stresses exceed the elastic limit of the geomaterial in the design.
When the stresses do not exceed the elastic limit of the geomaterial, the development of a
loosening zone is unaffected by the depth of the overburden. Hard rock is an example
here. On the other hand, when stresses fall in the plastic range of geomaterials, the effect
of overburden is a relevant factor, not only regarding the magnitude of pressure on the
underground structure but also regarding the time required for the development of a loos-
ening zone. Consequently, the use of methods taking the depth factor into account is nec-
essary. Crushed rock and sand are examples for this case. Therefore, we can divide the
existing empirical methods into two major groups, one considering the overburden depth
and the other neglecting it.
5.3 Methods Considering the Overburden Depth
5.3.1 Bierbiumer's Theory (1913)
Bierbiiumer's theory was developed during the construction of the great Alpine
tunnels (Sz6chy, 1973). According to this theory the tunnel is acted upon by the load of
rock mass bounded by a parabola of height h = ocH (Figure 5.1) in which a is the
"reduction coefficient". The rock mass in the pressure bulb of Figure 5.1 is called a
"loosening zone" which produces the load acting on the tunnel.
Bierbaumer proposed the following approach to estimate the reduction coefficient
a. This approach assumed that upon excavation of the tunnel, the rock material tended to
slide down along rupture planes inclined at (450 + 4/2). The configuration is shown in
113
Figure 5.2. The weight of the sliding rock mass was conteracted by the friction force (S)
developing along the vertical sliding planes, which can be expressed as:
S = fE = tan-.[tan 2 (450 4- H y HH
H2 y tan[ tan2(45 0 - ]
2 2 2
(Eq. 5.1)
in which
f= coefficient of friction = tan 4,
= friction angle of rock mass,
E = normal forces acting on the vertical sliding planes
= (area of triangle DFB' or ECA' in Figure 5.3) x y x K,
= [ .H Htan(450 -)]xy x tan(45 0 -

2 2 2
K,= Rankine's active ratio = tan(450 - /2),
y = unit weight of rock mass,
H = depth of tunnel (from the ground surface to the crown of tunnel).
All the formulas here are considered unit length in the direction perpendicular to the un-
derground opening face (i.e. plane strain condition). Therefore, the weight of rock mass
(W) acting on the tunnel crown (using a width B instead of b, because the assumption of
two vertical sliding planes in Figure 5.2) is:
W = y x area of C AB'D = y x H x B = yH[b+2m. tan(450 -)]

2
(Eq. 5.2)
in which
114
H = the height of rock mass above the tunnel,
B = the distance between two sliding surfaces = b+2m.tan (450 - /2),
b = the width of the underground opening,
m = the height of the underground opening.
Because of the shear forces on the sliding surfaces, the weight of rock mass above the
tunnel will not fully act on the tunnel. Hence, the force acting on the tunnel (P) is the dif-
ference between W (Eq. 5.2) and 2S (Eq. 5.1):
P = W-2S = yH[b +2m. tan(450 - tan4 [ tan 2(450 -D)]

y)]-H2
2 2
(Eq. 5.3)
The pressure on the crown of the tunnel (p) is thus equal to:
H tan tan2(450
PP - Hy[[1- 2]= aHy
B b+2mtan(450 - ) b +2mtan(45 0 - )
2 2
(Eq. 5.4)
in which
Htan tan2 (450 -

a = reduction coefficient = [1 ] (Eq. 5.5)
b+2mtan(45 0 -- )
2
Therefore only the pressure aHy, instead of Hy, acts on the tunnel. In other
words, h = aH is the effective height of rock mass above the tunnel that should be con-
sider when designing a tunnel. This implied the geostatic pressure was reduced by friction
produced by the horizontal earth pressure of the wedges ECA' and DFB' acting on the
vertical sliding planes. This pressure reduction was the so call arching effect.
115
Bierbaumer considered the force acting on the trap door calculated force from the
above method as "the maximum load". He proposed the other simplified method to calcu-
late the possible minimum load acting on the door. This method utilizes the equilibrium of
a triangular prism which has two sides making an angle 4 with the vertical (Figure 5.3).
When friction is fully mobilized along the sides, the forces on these sides have no vertical
component. "The minimum load" (Pi) on the tunnel roof, then just balances the weight
of the triangular prism:
Pm~ -Y (Eq. 5.6)

4 tan
Szechy (1973) discussed the Bierbiumer's theory. He mentioned that the correct-
ness ofBierbaumer's theory could not be completely verified in practice. The best results
were obtained for openings excavated at great depths in materials displaying high internal
friction (shear strength).
5.3.2 Balla's Theory (1963)
Balla (1963) assumed in his theory that the geomaterial lying above the tunnel
would suffer loosening and downward movement as a consequence of tunnel excavation.
This downward movement would take place along some kind of sliding surface and con-
sequently the displacement would suffice to mobilize the shear strength of the material.
He assumed the sliding surfaces were circular and they would start from the upper
corners of a rectangular tunnel (Figure 5.4). The radii of the circular curves were deter-
mined by the following procedure: (Balla, 1963)
116
1) Find an arbitrary point on a horizontal line at the elevation of the tunnel crown
(like point O1 in Figure 5.4).
2) Use the distance between 01 and the distant upper corner of the rectangular
tunnel as a radius to plot a circular curve (such as radius R in Figure 5.4).
3) Find the intersection of this circular curve and the axis of symmetry of the tun-
nel (point O in Figure 5.4).
4) Plot a straight line through the intersection point found in step 3) and tangent
to the circular curve made in step 2) (line OA in Figure 5.4).
5) Find the inclination angle (to the horizontal) of line OA (i.e. angle 0 in Figure
5.4) and the friction angle of rock mass ().
6) If the inclination angle (0) is equal to (450 - 4/2) which corresponds to that of
the sliding planes of passive earth pressure, then the radius R obtained in step
2) is the radius for the circular sliding surface. Figure 5.4 shows the case of 0
= (450-/2). Use the same radius and point 02 to plot the other sliding surface.
The area between these two sliding surfaces and the crown of the tunnel
(00102) is the loosening zone in rock mass.
7) If the value of the inclination angle is not (450 - /2), repeat step 1) to step 5)
until the angle is equal to (450 - 4/2).
The rock pressure on the tunnel is obtained from the equilibrium of all forces act-
ing vertically in and on the loosening zone (00102):
G + 2Qv + 2Kv + P = (Eq. 5.7)
in which
117
m
G = the weight of the loosening zone (or the sliding rock mass) enclosed
within the circular sliding surfaces,
Qv & K, = the resultants of shear stresses acting along each sliding surface,
P = the resultant rock pressure acting on the roof (crown) of the tunnel.
The decrease of load on the tunnel supports due to arching is taken into account in the
variables Q, and K,. Balla determined the distribution of stresses along the sliding sur-
faces (i.e. Qv and K, in Eq. 5.7) by Kotter's differential equation. He found that the vari-
ables in Eq. 5.6 were functions of the dimensions, locations and depth of the tunnel and a
function of the strength characteristics of the overlying rock. By assuming a uniform dis-
tribution of the crown pressure, he obtained:
b c (Eq. 5.8)
p.= Hy(FH+-Fb--Fc)
H Hy
inwhich
p, = the pressure on the roof (crown) of the tunnel,

H = the depth of the tunnel (from the ground surface to the crown of the
tunnel),
y = unit weight of rock mass,
b = the width of the tunnel,
c = cohesion in rock mass,
FH = factor representing the depth effect on p,,
Fb = factor representing the width effect on pa,
FC = factor representing the cohesion effect on p,.
118
FH, Fb, and F, are functions of the friction angle of rock mass (f). The values for these
factors can be found inFigure 5.5 provided by Balla. Some factor values are also listed in
a table in that figure. According to these numerical values, one can see that the effect of
cover depth (FH factor) is predominant.
In Balla's theory, many assumptions based on his experience were made. The de-
fect inthis theory is the fact that both the loosening and pressures of the lateral rock mass
beside the tunnel are neglected as the sliding surfaces start from the corner points.
5.3.3 Terzaghi's Recommendations for Load on Tunnel Supports (1946)
As discussed in Chapter 3, Terzaghi conducted the most widely known experimen-
tal and theoretical investigations of arching (Terzaghi, 1936 & 1943). He also combined
the knowledge and experiences obtained from many tunneling projects to produce design
values for tunnel support loads under various ground conditions (Proctor and White, 1946
& 1977). Many of Terzaghi's recommendations are still in wide use today.
Terzaghi thought the load on tunnel supports was the height of geomaterial which
tended to drop out of the tunnel roof (crown). This load on tunnel supports depended
more or less on the state of stress in the ground prior to tunnel construction. Also the
geological condition of the ground would affect the load on tunnel supports. In perfectly
or almost perfectly intact rock, no support was required unless popping was encountered.
In stratified or moderately jointed rock, the tunnel support served its purpose if it was ca-
pable to sustain a moderate rock load. In crushed rock, the loading conditions were simi-
lar to those to be encountered when mining through sand; and in zones of rock decom-
119
position, they were similar to those in tunnels through clay (Terzaghi, 1946). According
to these descriptions, Terzaghi suggested different values of load on tunnel based on the
ground conditions.
In this section, the focus is on the arch action that Terzaghi explored in his obser-
vations in construction and experiments, since it is the major topic of interest in this litera-
ture study. Within all of the different kinds of ground conditions, the crushed rock case
and the sand case assume an arching effect. These two cases will be discussed in detail in
the following paragraphs.
The term arch action indicates the capacity of the rock located above the roof of a
tunnel to transfer the major part of the total weight of the overburden onto the rock lo-
cated on both sides of the tunnel. The body of rock which transfers the load will briefly be
referred to as the groundarch (Figure 5.5). The arch action is the consequence of the lo-
cal stress loosening produced by tunnel excavation. The mechanics of the arch action is
shown in Figure 5.6. The shaded area (acdb) in that figure is the ground arch with a
width B1, and the zone of arching is assumed to have a depth D. While the tunnel is being
excavated and the support is being installed, the mass of crushed rock or cohesionless
sand constituting the ground arch tends to move into the tunnel. This movement is re-
sisted by the friction along the lateral boundaries ac and bd of this mass. The ground arch
shown in Figure 5.6 presents the concept of arching. However, it is inconvenient to use
this configuration in actual calculation. Terzaghi, therefore, provided a simplified model

to calculate the load on tunnel supports in Figure 5.8. The friction forces transfer the
major part of the weight of the overburden, with height H, onto the material located on
120
both sides of the tunnel. The roof support of the tunnel carries only the balance which is
equivalent to a height H,. Hp is the "equivalent overburden height" (Figure 5.8). This
height is used to represent the pressure on a tunnel in this section. The actual pressure
acting on the tunnel can be obtained by multiplying this height with the total unit weight of
the geomaterial.
Terzaghi suggested that the thickness, D, of the ground arch was roughly equal to
1.5B 1 while above the ground arch the pressure conditions in the rock remained practically
unaffected by the tunnel operations. Within the range of D (Zone of Arching, Figure 5.6),
a very small downward movement of the tunnel crown suffices to reduce the rock load on
the support of the intrados of the arch to a value HpIkm, which is much smaller than the
thickness, D, of the ground arch (Terzaghi, 1946). If the crown of the ground arch is al-
lowed to subside still more, the rock load on the roof support again increases and ap-
proaches a value Hp m, which is still much smaller than D (Terzaghi, 1946).
After the roof support is installed and tightly backpacked, the rock load increases
at a decreasing rate by about 15% from Hp to HI,t, i.e. Hpt = 1.15H , (Figure 5.7). The
load-depth relationship for a tunnel in sand or crushed rock is thus indicated by curve C in
Figure 5.7. The 15% increase of H, is the so called "time increase of the rockload". Ter-
zaghi also suggested that Hp could be related to the width B1, the width of the ground arch
(Figure 5.6), and that B1 was roughly equal to the width of tunnel (B) plus the height (Ht).
Therefore, the load on a tunnel, H,, is represented as follows:
HP= C x B1 = C x (B + Ht) (Eq. 5.9)
inwhich
121
C = a constant depending on the degree of compactness of the crushed
rock or sand and on the distance d through which the crown of the
ground arch subsided while the tunnel was constructed and the roof
support was being installed.
The distance d here is not known and it can hardly be determined by practical means.
Hence, the values for the constant C in Eq. 5.9 were recommended by Terzaghi from his
model testing results and experience. He considered different conditions of the tunnel: (1)
above water table or below water table, and (2) various geomaterials which include loose
sand, dense sand, and rock. The values for Hp are shown in Table 5.1. The ultimate value
in the table is equal to 1.15 times the initial value as discussed above. Terzaghi simplified
the real ground arch shown in Figure 5.6 to the configuration in Figure 5.8. Only the rec-
tangular part above the tunnel roof is carried by tunnel itself. The other parts are carried
by arching effect or by the rock wedges at the sides of the tunnel.
According to Terzaghi's experience, the actual roof load on a tunnel was normally
much closer to the minimum than to the maximum values (Table 5.1). This fact indicated
that a slight movement of the geomaterial towards the tunnel would activate the arching
q*et in the entire ground arch. He suggested a quick installation of tunnel support since
the load on a tunnel would increase from H mi. to H,.x gradually if more settlement of
the ground arch was allowed.
5.4 Methods Neglecting the Overburden Depth
122
The methods in this group essentially determine the loosening pressure in the geo-
material which is caused by underground excavation. Any relationship between the over-
burden depth and the pressure on the opening is neglected. A common feature of these
theories is the assumption that the temporary supports, or permanent lining of the opening
will be acted upon only by the weight of the stress-free body developed as a consequence
of loosening following excavation. Only one method will be introduced here.
5.4.1 Kommerell's Theory
The oldest and most widely known of these theories is that developed by Kom-
merell, who determined the height of the loading body from the deformations of the sup-
porting structure in the opening (Sz6chy, 1973). The theory is justified by the considera-
tion that the displacement, or deflection of the supporting structure is representative of the
displacement in the disturbed mass of geomaterial. As a consequence of this displacement
geomaterial is relaxed to a height h, which is equal to the height of the geomaterial column
capable of filling this space of height eby loosening. Hence, if the ensuing specific strain
is 8%, then:
h8
e = h(Eq.
100 5.10)
100-e
and thus h= (Eq. 5.11)
8
in which
8 = the loosening coefficient of the geomaterial as a percentage. Some
loosening coefficients for different geomaterials are given as follows:
123
loose granularsoil (sand) 8 (%) = 1-3 %
moderately cohesive soil (dry clay) 6 (%) = 3-5 %
cohesive soil (marl,gravely clay) 8 (%) = 5-8 %
soft rocks (sandstone, limestone) 8 (%) = 8-12 %
solid rocks 6 (%) = 10-15 %
The tunnel support is required to carry the weight of a rock mass bounded by a
half ellipse of height h, as shown in Figure 5.9. Point O is the point of origin in the coor-
dinate system, x and y are coordinates of the points on the ellipse, e is the deformation for
the geomaterial, h is the height of the loosening zone, and b is the width of the tunnel.
Therefore, the equation of the ellipse in Figure 5.9 will be:
X22 2
=1 (Eq. 5.12)
(b/2) 2 h2
Substituting h in Eq. 5.12 with the h value in Eq. 5.11:
4x 2
4X2 + y262 =1 (Eq. 5.13)
b2 10000e 2
The total load on the tunnel is:
P = y x (Area of the half ellipse)

1 b 1 b 100e
y x(n- h) y x-(7 (Eq. 5.14)
2 2 2 2 8
25yirbe
Therefore, if we know the loosening coefficient (8) from the type of the geomaterial, and
assume the permissible deflection of the tunnel roof (e), we can obtain the height of the
loosening zone (h) from Eq. 5.11, the shape of the loosening zone from Eq. 5.13, and also
the total load on the tunnel roof (P)from Eq. 5.14.
124
Nevertheless, Kommerell's theory can only be used as a rough approximation be-
cause of the following reasons:
1) Loosening is possible in granular soils or crushed rocks as a result of the redis-
tribution of individual particles. Loosening is impossible in bulk solid rocks,
where only elastic expansion upon load release of the order of a few millime-
ters can occur.
2) The linear relationship assumed to exist between the deflection of the roof (e)
and the height (h) of the loading mass could not be verified by measurements
nor explained theoretically because it is purely an empirical relationship.
5.5 Discussion
The empirical and semi-empirical methods are often applied by engineers in con-
junction with other approaches, such as numerical analyses, to the design of underground
structures. Coefficients used in these methods may include the conditions (like depth) and
the properties (like friction angle) of the geomaterial. Generally, these methods only pres-
ent the pressure acting on the support system of a buried structure and not the displace-
ments of the geomaterial and structure. The benefits of using the empirical methods are
"quick" and "simple". If engineers choose a method which was derived from experiences
with ground conditions and construction methods similar to those for the proposed struc-
ture, good results can normally be obtained.
However, there are still some uncertainties in these empirical methods. Especially
some of the assumptions in these methods may not be realistic, e.g. the assumptions of the
125
loosening zone in aforementioned sections ,and the simplified equilibrium equations for the
forces inthe loosening zone. In addition, some assumptions are even incorrect under cer-
tain circumstances. For example, the theories in Section 5.4 which deal with the determi-
nation of loosening pressure in rock masses due to underground construction, neglect the
effect of depth. These theories assume the development of arching action in the geomate-
rial, and consider that the size of the load mass on the top of a tunnel only depends on the
size of the opening and the strength characteristics (such as the internal friction angle of
the geomaterial). These assumptions are invalid when the opening is close to the ground
surface or at a small depth. They are also invalid when the geomaterial undergoes plastic
deformation or time dependent behaviors ,since the depth of the geomaterial will be an
important factor under these circumstances. Comparing the empirical methods, based on
the previous assumptions, with other analytical methods, suggests analytical methods are
more appropriate in actual situations as they attribute a decisive role to the manner in
which the stress distribution is changed in the vicinity of the opening. Hence, empirical
methods should be applied in conjunction with the other analytical methods, in order to
confirm their results.
126
Table 5.1 Overburden Load (in feet) in Sand and in Blocky and Seamy Rock (Proctor and
White, 1946)
Above water table Below water table'

Material
Hrp min Hp max Hp min Hp max
Dense sand" Initial 0.27 (B+ H,) 0.60 (B + -H,) 0.54 (B H,) 1.20 (B+ H,)
Ultimate 0.31 (B-4-H,) 0.69 (B + H,) 0.62 (B H,) 1.38 (B+ H,)
Loose sand" Initial 0.47 (B - H,) 0.60 (B -- H,) 0.94 (B H,) 1.20 (B + H,)
Ultimate 0.54 (B4- H,) 0.69 (B- H,) 1.08 (B + H,) 1.38 (B+ H,)
3
Moderately blocky HHn= 0 increasing up to H,,,, = 0.35 (B+ H,)
Very blocky and shattered H., , = .60 (B+ H,) increasing up toHP,, = 1.10 (B+ H,)
.. Values are rouqhay equal to twice those for dry sand.
2. Values computed on basis of laboratory tests.
3. Values computed on the basis of the results of observations in railroad tunnels.
127
I
Ground surPace
-4------n
Figure 5.1 Rock Pressure Bulb after Bierbaumer (Szechy, 1973)
128
- B J
2S -lfe -2IM. lanc (05 --
Figure 5.2 Assumption Model of Bierbiumer's Theory: Maximum Load

(Szichy, 1973)
~ ~T~- J ~~TTR-- ~-'-~-~T~~~.r ~T---~- -;-T-.T ~---~--- 7~7r ~- - ~ --
B_
Figure 5.3 Assumption Model of BierbAumer's Theory: Minimum Load
(from Iglesia et. al., 1990)
129
Figure 5.4 Principle of Balla's Theory (Balla, 1963)
Fr
I.j
:.3
"9
t? 7' -7 47
Figure 5.5 Resistance Factor Diagram and Table

(Balla, 1963)
130
Surface
7
tI I,
1
D
(Zone of
arching)
1
r
1
- B-
Figure 5.6 Configuration of Ground Arch (Proctor and White, 1946
131
0.15 H,,
HI,
Rockload in feet
Figure 5.7 Load-Depth Relationship on a Tunnel in Sand or Crushed Rock

(Proctor and White, 1946)
132
_ Sand surface
Carried by arching
Approx. B + H
t
*1 Cc
by
df
Figure 5.8 Simplified Model of Load on Tunnel Support (Proctor and White, 1946)
133
A B
Figure 5.9 Shape of Kommerill's Pressure Diagram; Line AB: the Tunnel Roof
(Sz6chy, 1973)
134
Chapter SIX
Experimental Investigations and
Photoelasticity Methods
6.1 General
In this chapter, several experimental investigations will be reviewed. After Ter-
zaghi's trap door experiment, several different trap door experiments were performed.
They are similar to Terzaghi's experiment but with different improvements or extensions.
For example, McNulty (1965) increased the pressure on the trap door to simulate the in-
situ stress level. Ladanyi and Hoyaux (1969) used an ideal granular mass (aluminum rods)
in plane strain in their trap door experiments to check the validity of the classic bin theory
(i.e. the assumption of vertical failure planes in Terzaghi's arching theory, see Section
3.2). Harris (1974) performed experiments with several trap doors to investigate the ef-
fect of advancing face in underground construction. Evans (1984) used different shapes
(circular and rectangular) of trap doors to conduct his experiments. Iglesia et. al. (1990)
performed the trap door experiments with different materials (sand, wooden, and alumi-
num rods) in a centrifuge to explore the arching effect. Photoelastic methods used in ex-
ploration of stresses around underground structures (Riley, 1964) and interparticle contact
forces (Paikowsky et. al., 1993 & 1995) will also be discussed in this chapter.
6.2 Experiments by McNulty (1965)
The trap door experiments from Terzaghi (1936) are small scale experiments. The
stresses acting on the trap door are fairly small comparing to those stresses in real con-
135
struction sites. In order to reach stress levels commonly encountered in the field, McNulty
(1965) applied air pressure at the surface of the sand in a trap door experiment which is
otherwise similar to Terzaghi's 1936 experiment.
In McNulty's experiment, a cylindrical soil chamber with a circular trap door at its
bottom was used (Figure 6.1). McNulty developed both active and passive arching in his
study by measuring the pressure acting on the trap door (PB) as the door displaced upward
or downward by an average displacement (8). In addition, the effects of overburden depth
(H), trap door diameter (B), surface pressure (Ps), and different soil properties were also
evaluated in his study. The layout of McNulty's axisymmetric tests is shown in Figure 6. 1.
There is a parameter called the "arching ratio" (AR) in Figure 6.1. This parameter equals
the pressure on the trap door divided by the surface pressure:
AR (arching ratio) = PB(pressue on the trap door) (Eq. 6.1)

Ps (surface pressue)
In order to investigate the influence of different soil properties, two sands were
used. Sand 1 was the Reid-Bedford Model sand. It was a clean, uniform, find sand with
D0o = 0.16 mm and average friction angle of 330. Sand 2 was the Cook's Bayou sand. It
was a clean, uniform, medium sand with Dio = 0.22 mm and average friction angle of 380.
Typical results from McNulty's series of axisymmetric tests are shown in Figure
6.2 (sand 2 is used here). These results show that even a very small deflection (i.e. a very
small value of 8/B in the figure) can cause considerable changes in the load on the trap
door. This outcome is similar to the previous observations by Terzaghi (1936). However,
in Terzaghi's experimental results for active arching, the pressure on the trap door would
increase gradually and reach a constant ultimate value (i.e. a normalized force,
136
F/Foxl00%, which is about 12.5%) after a sudden decrease (F/Fox100% = 6% ~ 9.6%), at
the beginning (details in Figure 3.2). This phenomenon was not observed here. The load-
displacement curves for active arching in Figure 6.2 do indicate that there was an apparent
decrease of pressure, but the pressure on the door did not increase afterward; it (pressure)
would keep decreasing until an ultimate value was attained.
The load-displacement curves in the active arching cases show that the load reduc-
tion is greater as the depth of soil overburden increases. For the passive arching (i.e. the
trap door moving upward), the increase in load is greater as the overburden depth in-
creases (e.g. the H/B =2 curve in Figure 6.2). However, McNulty found that the influence
of H/B value on the passive arching curves was minimal when H/B > 2. This indicates
that as the ratio of the soil depth to the trap door diameter (H/B) increases in the active
arching case, a stronger soil arch will be developed in the soil body and transfer more load
from the trap door. On the other hand, the trap door will take more pressure when the
soil depth/trap door size ratio increases (i.e. H/B is greater) in the passive arching case.
McNulty also compared results of the two different types of sands he used in his
experiments. Sand 2 was stronger than sand 1 since sand 2 had a higher friction angle
(380). The results for the active arching experiments with a surface pressure of 75 psi are
depicted in Figure 6.3. For the shallow soil depths (H/B = 1/3 and 2/3 in Figure 6.3), the
ultimate arching ratios differ by about 10%, though the trap door moves the same for both
sands to develop the ultimate arching ratio. Sand 2 has the lower value in the figure and
this means it can transfer more load than sand 1. At H/B = 4, the ultimate arching ratios
of these two sands are close to each other, but it takes sand 1 twice the door movement to
137
reach this ratio than sand 2 (see the bottom plot in Figure 6.3). The similar behavior of the
two sands at small values of H/B is attributed to the immediate attainment of a state of
plastic equilibrium caused by a slight door movement with shallow soil depths. The differ-
ence at greater soil depth is due to the difference in the soil strength (i.e. the friction an-
gle).
6.3 Experiments by Ladanyi and Hoyaux (1969)
Ladanyi and Hoyaux (1969) performed an experimental program involving an ideal
granular mass overlying a model trap door in plane strain conditions. Figure 6.4 shows
their apparatus. The geomaterial was simulated by a stack, 40 inches high and 80 inches
wide, of aluminum rods supported by a U-shaped rigid steel frame. The rods were 2.5
inches long and had circular cross sections with two different diameters, 1/8 inch and 3/16
inch. They were cut to the required length, sanded in a sand drum and then mixed in equal
proportions to form the testing samples. The resulting granular medium had a unit weight
y m 0.079 lb/in3 and a friction angle 4 m 290 at the anticipated stress levels. No surface
pressure was applied. Hence, the anticipated stress levels in the experiments were very
low. The trap door here was a buried structure, 3 inches wide and 2.5 inches long, repre-
sented by a rectangular rigid metallic box, which could be moved up or down like a piston
inside the big testing box with aluminum rods. Figure 6.5 shows the downward (Figure a)
and upward (Figure b) movements of the trap door structure.
In the tests, the pressure on the trap door was measured as a function of the verti-
cal door movement. The rod displacement trajectories were photographically recorded.
138
By painting a six-inch square mesh grid on the rod sample, a photo of the overall defor-
mation pattern of the sample was taken after the trap door was moved a certain amount.
Figure 6.6 shows the results when the trap door had undergone a large downward move-
ment (3.6 inches). It was observed that the deformation of the grid (Figure 6.6) was well
limited to a narrow band above the trap door. Some lateral movement of the granular
material toward the centerline was also observed during the test. This was because the
granular material tended to fill the space generated by the downward movement of the
trap door. The results obtained from four active arching simulations are shown in Figure
6.7. Four different overburden depths were used. They were 6 inches (H/B = 2.00), 8.5
inches (H/B = 2.84), 13 inches (H/B = 4.33), and 16 inches (H/B = 5.33). The maximum
displacement of the trap door in these tests was 4 inches. The results, illustrated in Figure
6.7, were similar to the results that other researchers had observed. A rapid drop of pres-
sure on the trap door occurred shortly after the door was lowered. In the cases shown
here, the minimum pressures were obtained after the door moved downward of about 8%
of the trap door width (Figure 6.7). With further door movement, the pressure on the trap
door increased gradually until an ultimate value was reached. The reduction in pressure
due to active arching became more significant as the soil overburden depth became greater
(i.e. H/B is greater).
A theoretical approach used to estimate the pressure reduction on the trap door
was derived by Ladanyi and Hoyaux. However, the major purpose of Ladanyi and
Hoyaux' was to examine the appropriateness of the assumption of two limiting vertical
failure planes in the traditional arching analysis (Terzaghi, 1943). A comparison of meas-
139
ured and calculated vertical pressures for a downward moving structure (active arching) at
different depths of burial is shown in Figure 6.8. The calculated results were close to the
measured results. Hence, Ladanyi and Hoyaux believed that, within the depth interval in-
vestigated, the variation of pressure acting on a vertically translating rigid horizontal trap
door could be reasonably predicted by assuming the existence of two limiting vertical fail-
ure planes extending from the edges of the trap door to the free surface (see Figure 3.5 &
Figure 3.6). A correct pressure prediction was, however, only possible when proper con-
siderations were made concerning the lateral stress ratio (K) and the value of normal
stresses (Oh) acting along the assumed vertical failure plane.
6.4 Experiments by Harris (1974)
Harris (1974) performed sandbox trap door experiments aimed at simulating the
stress redistribution around a long wall coal mining operation. Harris' apparatus is shown
in Figure 6.9. It consisted of a series of trap doors which could be lowered independently
to model the advancing face. Diaphragm pressure cells were located as shown to measure
the stresses on the trap doors. By lowering the doors in succession, Harris obtained the
stress distribution from beyond the influence zone ahead of the face. Typical results are
shown in Figure 6.10. Areas of stress concentration (at abutments) developed ahead of
and to the side of the face. A less developed abutment (the "rear abutment area" in Figure
6.10) appeared within the extracted zone (above the lowered doors). Harris' results can
also be interpreted as the stress distribution around an advancing tunnel face in soft
ground. We can conclude that the stress field near the face is not like that predicted by
140
most approaches. Instead, it is a very complicated three-dimensional pattern. A constant
stress distribution does not develop until one to two face widths behind the advancing
face. Similar experimental results were found in Evans' experiments (details in Section
6.5).
6.5 Arching in Granular Soil (Evans, 1984)
6.5.1 Experimental Setup
The theoretical continuum approach using plasticity theory developed by Evans
(1984) was discussed in Chapter 4. This section describes several trap door tests that
were performed by Evans to verify the validity of the theoretical approach and to further
explore the load redistribution process in buried structures.
Evans' trap door tests included: (1) active arching above a circular trap door, (2)
active and passive arching above a rectangular trap door under plane strain conditions, and
(3) active arching above a row of trap doors lowered in succession so as to simulate an
advancing tunnel. Figure 6.11 shows the circular trap door apparatus which was an initial
test instrument. Evans constructed it for the purpose of determining whether the dia-
phragm type pressure transducers would give accurate results in his arching study. Figure
6.12 presents the rectangular trap door apparatus which was the primary test instrument in
his study (plain strain conditions). The rectangular trap door apparatus was installed with
a series of rectangular trap doors so it also can be used for the simulation of an advancing
tunnel. Basically, the instruments used in Evans' experiments were similar to those used
in the other trap door tests (e.g. Terzaghi's test in 1936). Evans, however, used the dia-
141
phragm type transducers to get more accurate pressure readings on the trap doors and on
the bottom of the testing box.
In order to investigate sand type effect, four different sands were used: (i) Fine
Leighton Buzzard sand, (ii) Coarse Leighton Buzzard sand, (iii) medium tan sand (Bin 23
sand), and (iv) fine white sand. The soil samples were prepared by the raining deposition
technique, resulting in loose-sand samples.
6.5.2 Test Results
Figure 6.13 shows the variations of soil displacement pattern with increasing trap
door displacement during an active arching test in the primary test apparatus. A triangular
shaped zone expanded vertically with noticeable dilation present. This result matches
Evans' analytical approach by plasticity theory which was shown in Figure 4.9 (a). When
the dilation angle (v) is larger than zero, the soil body expands and the free body diagram
is triangular. The relation of normalized vertical stresses on the trap door versus its dis-
placements is shown in Figure 6.14. This relation is, again, similar to Terzaghi's trap door
testing results (Figure 3.2, Loose sand) and typifies the active arching behavior observed.
Vertical stresses decreased rapidly as door movement began, until they reached a mini-
mum value. As door motion proceeded, the stresses went up to a point where the dis-
placement was about 10% of the trap door width, at which the stresses became quite
steady afterwards.
As to the passive arching case, Figure 6.15 shows the typical patterns of soil de-
formation in Evans' tests. A trapezoid zone expanded in the vertical direction and showed
142
apparent dilation in the soil body. Evans' plastic analytical model for passive arching in
plane strain condition is similar to this result when v > 0 (Figure 4.10 (a)). Figure 6.16
describes the quantitative results of Evans' passive arching tests. The stresses on the trap
door increased quickly as soon as the door was moved upward and then reached a peak
value. After the peak, the stresses on the door began to decrease till a fairly constant
value was reached. The passive arching test results of Evans' experiments are similar to
McNulty's results shown in Figure 6.2, differing by the fact that no apparent peak value
was observed in McNulty's results. McNulty found that the ratio of the overburden soil
depth (H) to the trap door width (B) affected the level of arching obtained. Evans' test
results confirmed that observation. The percentage change in the trap door stresses in-
creased as the H/B ratio increased (Figure 6.2 and 6.16).
The simulations of an advancing tunnel by successive lowering of a series of trap
doors were performed by Evans' study as well. When the trap doors were lowered (active
arching) or raised (passive arching) in consequence, the deformation of the soil body was
observed (as shown in Figure 6.17 and 6.18) and the stresses on each trap door were
measured. Evans found that, even at shallow depths, the "tunneling" process caused redis-
tribution of stresses more than one diameter in advance, at least 0.5 diameter to each side
and 1.5 diameters behind the face. His results clearly showed that those arching ap-
proaches which do not account for three-dimensional behavior will incorrectly model the
kinematical compatibility in soil. Evans believed that instead of no volume change as as-
sumed in the traditional two dimensional arching models, there should have been a volume
increase inthe soil when arching occurred.
143
The use of four different sands in Evans' tests resulted in similar behaviors except
in the tests simulating an advancing tunnel. In these tests, smaller stress redistribution was
observed for the fine white sand compared to the medium tan sand under identical test
conditions. However, no further explanation was given in Evans' study.
6.5.3 Comparison of Plasticity Theory Approach and Test Results
The theoretical results presented in this section are based on the primary test appa-
ratus in a plane strain condition (Figure 6.12). Evans compared his test results with the
predicted values based on his plasticity approach (Section 4.3) and those developed oth-
ers. A dimensionless "load factor" (CQ) was used for the comparison shown in this sec-
tion. It was defined as:
AV
cc =- -2 (Eq. 6.2)
PfB2
inwhich,
F : the force per unit length of door parallel to the infinite axis, i.e.
F= B-(av),,
y : the soil's unit weight,
B : the trap door width.
The value of the coefficient of lateral stress (K) needs to be decided before using
Evans' theoretical approach to calculate the force acting on the trap door. Evans has ac-
tually measured the K value in some of his experiments. He noticed its variability depend-
ing on the vertical location as well as the level of trapdoor displacement. He picked out K
equal to 1.2 to calculate all of the theoretical results.
144
(I) Plane StrainArching/Active Case:
Figure 6.19 and 6.20 provide comparisons between experimental results and pre-
dicted values through plots of load factor (C) versus H/B. Figure 6.19 presents experi-
mental results for the maximum arching points (the peak) from Evans' tests, Terzaghi's
tests (1936), and Harris' test (1974). Several theoretical predicting results are also shown
in Figure 6.19. They include Sz6chy (1966), Vertical Slip Surfaces (Matyas & Davis,
1983a), Silo Theory (Janssen, 1895; Jakobson, 1958), Silo Theory as Modified by Ter-
zaghi (1943), and Evans' Plasticity Theory Solution (1984). Table 6.1 has a summary of
the formulas for the aforementioned theoretical methods. All the curves are drawn for the
sand friction angle () = 350 and K = 1.2. The plasticity theory solution is also shown for
additional friction angles. Figure 6.20 shows results at the ultimate arching state (the
steady value in Figure 6.14) and the predicting results from the plasticity theory solution.
A comparison of these two figures suggests: (i) better predictions of load factors at maxi-
mum arching than any of the existing theories, and (ii) reasonably good predictions of load
factors at ultimate arching. Since the soil body is in a plastic condition at ultimate arching,
Evans' approach is the only method able to be applied.
(II)Plane StrainArching/Passive Case:
Figure 6.21 and 6.22 provide the comparison between experimental results and
predicted values. Test data for the maximum arching state (Figure 6.21) are taken from
Matyas and Davis (1983b) as well as Evans' own tests. Different theoretical approaches
145
were used in Figure 6.21 including US Standard for Rigid Pipe (Matyas and Davis,
1983a), Vertical Slip Surfaces (Matyas and Davis, 1983a), Ladanyi and Hoyaux (1969),
Das and Seeley (1975), and Evans' Plasticity Theory Solution (1984). Table 6.2 lists the
formulas of the aforementioned theoretical methods. All the curves are drawn for the sand
friction angle ( ) = 350 and K = K. (Active Rankine Ratio, K.=- sin ). Evans' theory
1+ sinO
once again provides the best prediction of forces at maximum arching. However, it con-
sistently overestimates the magnitude by 10 to 15% (Figure 6.21). Evans explains that
this overestimation is due to the assumption in his theory that uniform stresses across a
trap door exist during passive arching. As to the ultimate arching state, there are not
enough test data in Figure 6.22 to draw a strong conclusion regarding agreement with
predicted values.
6.6 Centrifuge Modeling of Jointed Rock (Iglesia et. al., 1990)
6.6.1 Introduction
When a model is constructed smaller in size than the prototype but uses the same
material as the prototype, the model is proportionately lighter than the prototype. This is
usually not a major problem in models where the dead weight of the system can either be
ignored or simply included in the external loads, for example, with applications in struc-
tural engineering. However, for geotechnical problems, the body forces can neither be
neglected nor replaced with equivalent external forces that will produce the same effects.
Therefore, if a geotechnical system is to be modeled physically at a small scale, both the
146
magnitude and the variation of geostatic stresses have to be duplicated in the model
(Iglesia et. al., 1990). The best way to do this is by using a centrifuge to attain the desired
geostatic stresses in the model.
Centrifuge modeling was started in the 19th century when a French engineer Phil-
lips(1869) suggested that a centrifuge be used to simulate self-weight stresses in structural
beams. Through the years, physical modeling of geotechnical systems with the centrifuge
has become very popular. Centrifuge modeling for geotechnical research has been ac-
cepted widely in soil engineering. Iglesia et. al. (1990) tried to find out whether the centri-
fuge modeling scheme also applied to jointed rock mass. The following section illustrates
their study.
6.6.2 Trap Door Experiments Using Centrifuge Modeling
The experimental setup adopted by Iglesia et. al. (1990) at MIT involved a jointed
rock mass model with a trap door underneath. The trap door design was similar to several
former experiments, e.g. Terzaghi (1936), McNulty (1965), Ladanyi and Hoyaux (1969),
Evans (1983), etc. Figure 6.23 shows the concept of the trap door system and Figure
6.24 presents the real trap door utilized in the centrifuge modeling. The whole rock mass
model system was spun up gradually to a desired gravity level, at which the trap door was
lowered (i.e. active arching case), and measurements of the force on the door with the
corresponding displacement were taken. Different sizes of small-scale models and of trap
doors were used with various levels of pseudo-gravitational acceleration in the centrifuge
at which the door was moved down. The primary objective of this research was to inves-
147
tigate the scaling relationships for centrifuge models of jointed media (wood rods and
aluminum rods). A series of tests with granular materials (coarse sands and glass beads)
was performed first to give an assessment of similitude for the designed system.
6.6.3 Experimental Results
A "modeling of models" exercise has been carried out with granular materials us-
ing various corresponding scales of particle size, door width, overburden depth, and grav-
ity level. The test results were served as a benchmark upon which the experiments of
scaling injointed models could be based. Figure 6.25 presents the trap door test results of
coarse sand with different soil depth. These curves showed similar shapes of the other
active trap door test results (e.g. McNulty, 1965) and had the same trends as soil depth
was increased, i.e. the deeper the soil, the greater the percent reduction in the load as the
trap door displaced. Iglesia et. al. also performed different tests by changing different
control factors. They found all of the results from various scaled models of the granular
medium showed good agreement between the model and the prototype. In particular, the
measured minimum loads on the trap door yielded the same average stress in appropriately
scaled setups. They reached a conclusion that the criteria for similitude under plane strain
conditions were satisfied inthe centrifuge trap door models with granular media.
The centrifuge tests were then conducted with jointed media. Figure 6.26 presents
an example which is the result from four sets of tests with different stack methods (direct
stack and brick stack) used to determine the effects of g-level (40g and 80g). Figure 6.27
shows the final configuration of one set (GI 192) of the tests in Figure 6.26. An arch can
148
be seen clearly at the bottom of the picture. However, the test results (load-displacement
relations) of jointed media did not completely fulfill the similarity requirements for appro-
priately scaled arrangements. The force value, especially with minimum loads on the trap
door, approximate satisfying the similitude criteria for all the scale considered. These
forces, nevertheless, occurred at about the same absolute door displacement which vio-
lated the centrifuge scaling laws. Similitude requirements in connection with the scaling of
length dimensions were, thus, not met by the observed displacements. Iglesia et. al. be-
lieved that this finding did not suggest that centrifuge modeling was totally useless for
situations involving jointed media, as they stated "The ability to generate body forces in
the centrifuge makes it an invaluable tool in obtaining data against which analytical and
numerical methods can be validated". They suggested, however, not to extrapolate results
from model tests to actual prototype conditions when dealing with the jointed media.
6.7 Photoelastic Methods
6.7.1 Introduction
It is difficult to use traditional stress measurement methods to measure stress dis-
tributions around underground structures of complex geometry. However, photoelastic
methods can solve this difficulty. Photoelastic methods are useful for measuring stresses
and strains. They are based on a phenomenon that occurs when polarized light passes
through certain homogeneous materials such as glass. Under straining, a brilliantly col-
ored fringe pattern appears in the material. This colored fringe pattern results from
149
changes in the index of refraction (ratio of the velocity of propagation of light in a vacuum
to its velocity in another medium) proportional to the amount of strain in a material in any
given direction. Using this relationship, a proportionality constant can be developed that
relates the "colored bands" to different levels of stress and strain.
In early photoelastic methods, stresses were induced in an elastic material which
was subsequently set, freezing the stresses in place. The elastic material was then sliced
and analyzed under polarizing filters. Because the material could be used only one time in
each test, the cost of conducting a test by photoelastic methods was expensive. The new
photoelastic method uses a special technique which can generate fringes in the material
when the material is subject to external loads, but will not freeze fringes in the material
after the external loads are removed. The fringes generated during the test are recorded
by a high speed camera with a polarized filter. This new method allows one to reuse the
photoelastic material and is, therefore, more economical.
6.7.2 Photoelastic Study in Embedded Structural Elements (Riley, 1964)
Riley (1964) used the new photoelastic method to conduct a series of studies to
determine stresses in the free field of a two-dimensional plate and on the boundaries of
embedded structural elements in the plate during passage of a stress wave. Two major
parts were included in his experiment. First, was the explosive study. The model used
was machined from a large sheet of low modulus urethane rubber. A 5/8 inch diameter
hole was machined 4 inches from the point of load application along a radial line 300 from
the centerline of the plate. The layout is shown in Figure 6.28. An explosive charge was
150
applied on the plate top at the centerline (Figure 6.28). A complete photoelastic fringe
pattern record was obtained using a camera operating at 6780 frames per second through
a polarized filter. Figure 6.29 shows an example of the fringe pattern around the hole at
1050 microseconds after the explosion. The stresses around the hole were then obtained
by transforming these different colored fringes to the stresses they represent. Figure 6.30
shows the static and dynamic stress distributions on the hole boundary 1050 microseconds
after the explosive charge was detonated. The static stress distribution was solved under a
theoretical approach which assumed that a static load equal to the explosive load was
applied on the plate top. The dynamic stress distribution (Figure 6.30) was transformed
from the fringe pattern shown inFigure 6.29.
The second part of Riley's experiment was the air shock study. The model used
for this part is shown in Figure 6.31. A 3/4 inch diameter inclusion was placed in a pho-
toelastic plate at the point which was on the centerline and 4 inches from the top of the
plate. The same material was used as in the first part. The air shock loading was applied
to the top edge of the model by means of a 6 inch diameter shock tube. Several fringe
patterns around the inclusion were obtained after the test.
Riley then compared the results of these dynamic test results with the static solu-
tions (like Figure 6.30). He concluded that static solutions could be used as a first ap-
proximation for computing dynamic stress distributions on the boundaries of discontinui-
ties in elastic materials. He suggested the photoelastic methods were useful in the study of
stress distribution around underground structures under static loads and propagating stress
waves.
151
6.7.3 Photoelastic Study in Direct Shear of Granular Material (Paikowsky,
1993 & 1995)
The same photoelastic technique was used by Paikowsky et. al. (1993) to study
interparticle contact forces. In their experiment, particles were modeled two-
dimensionally as circular and/or elliptical discs made from photoelastic material. This
makes the measurement of contact forces (direction and magnitudes) possible, and allows
particle tracking to evaluate their motion (translation and rotation). Riley's photoelastic
measurement technique required detailed measurements to determine the stress field in a
single plate. His method treated the whole testing system (a plate contains a hole or an
inclusion) as one piece. In other words, the plate and the hole (or the inclusion) in the
plate constitute a "continuum". However, if the interaction between discrete particles of a
granular material (e.g. sand or rock) is desired under different loads, the use of photoelas-
ticity for granular material modeling (Paikowsky et. al., 1993) would be more appropriate
than Riley's continuum system.
Paikowsky et. al. (1995) used their granular material modeling system in a two-
dimensional direct shear test. The direct shear box and samples are shown in Figure 6.32.
All the circular particles in Figure 6.32 were made of photoelastic material. A vertical
load was applied at the top of the whole particle system and the bottom solid surface was
pulled out toward the left in Figure 6.32. The particle displacements and interparticle
forces were measured when the bottom solid surface was being pulled out. Figure 6.33
shows the configuration of the particle system when the bottom solid surface was pulled
152
out by one particle diameter. All the particles in the second row (counting from the bot-
tom) rotated 600 clockwise. An "arch" was formed at the right hand side, bottom corner,
just like the soil arch observed in a sand direct shear test. This result demonstrated the
possibility of using this experimental system to investigate all the interactions between
particles, e.g. arching. However, not much qualitative analysis had been performed on the
data collected to date. More analyses of the experimental data are necessary to prove the
validity and versatility of their experiment. Further study in their experiment includes the
effect of particle shape and particle size to the mechanical behaviors of the whole particle
system, and a series of trap door tests to explore the arch effect in the granular material.
153
Table 6.1 Summary of Plane Strain Active Arching Formulae (Evans, 1984)
Szichy (1966)
Cc --(1 Hanotan2(450 - )) for H < 5
C= (tan 4(450- )) for 1

B 5
Vertical Slip Surfaces (from Matyas and Davis, 1983a)

H
Cc (I - KOB tan$) with KO= 1 - sino
Silo Theory
-2K H tanK)
C = 1 (1
c 2KtanO - e B
Silo Theory as Modified by Terzaghi (1943)

2K 1C
C = c 2Ktan
Ka 1- e) -2KH tanO H < 2
for B
C =
C- 2Ktang
1 (1 -4KtanO) + - 2) e-4Ktan 0
for iB >2
Plasticity Theory Solutions:

For Maximum Arching Point
C= H(1- tano) for w 1
< 1
1 B 2 tan1
c 4tanj
7
-,or B- > 2-tlnp
For Ultimate Arching State
Cc= 2Ksin f -< 2

for H
2KsinX
- 4
= i- Ksin$)
+ (H _ 2 )e- Ksinf
2e
4 for B- > 2
2Ksino 3
154
Table 6.2 Summary of Plane Strain Passive Arching Formulae (Evans, 1984)
U.S. Standard for Rigid Pipe (from Matyas and Davis, 1983a)
H
Cc= 1.961 H - 0.934
Vertical Slip Surfaces (from Matyas and Davis, 1983a)

Cc= H
!(Ko H
tans + 1) with K o = 1 - sins
Ladanyi and Hoyaux (1969)

C = H
H(+ H sin20)
c 2B
Das and Seeley (1975)
cc= H( Katano + 1)
Plasticity Theory Solutions --
For Maximum Arching State

H H
C = l(l + 1tanX)
For Ultimate Arching State
1
Cc= 2 Kasin (*.
e2 a sinH H
- 1) for
B -< 2
C= 1
2 Kasin
(e4 Kasin 1) H - 2) e4Kasins
c B
w 1 - sinO for H >2
where Ka 1 + sino
155
IA L I f I
II t I f
****ee*
********
************ -5
"" "" "
, * o- - -
........... :
- -e - * X**
- -
I a +6
AXIALLY SYMMETRIC TESTS

P
AR =- PS = 75 PSI
P
Figure 6.1 Layout of McNulty's Experiments (McNulty, 1965)
156
vt
0
o
e m o a o 0
d
-
3
'OIlY ol
,f acdl
cz
a,
C1
,,,
157
U
I* I I I I
LEGEND
0 SANo P SP
*
6 SAND3
i , ~-_I_ .I Ia
o I r .4
U
I
I.
i
I - I
I-~--~
I
I.'
U91
I
I g
I I
I I I I
Figure 6.3 Influence of Soil Properties and Overburden Depth on Active Arching
(McNulty, 1965)
Figure 6.4 Ladanyi and Hoyaux's Experimental Setup (Ladanyi and Hoyaux, 1969)
158
Figure 6.5 View of the displacement trajectories for (a) downward, and (b) upward
movement of the structure (Ladanyi and Hoyaux, 1969)
159
Figure 6.6 Deformed Square Grid after a Large Downward Movemen
(Ladanyi and Hoyaux, 1969)
U)
Cn
O
SETTLEMENT RATIO 6/B
Figure 6.7 Normalized Pressure (a/'yH) versus Settlement Curves oi

(Ladanyi and Hoyaux, 1969)
160
0
0
o
I-
t
w
Ja.
0-
W
PRESSURE REDUCTION RATIO(I-0vD)
Figure 6.8 Comparison of Measured and Calculated Vertical Pressures for a downward
moving trap door at different depths of burial (Ladanyi and Hoyaux, 1969)
161
-mCNTIU
*ttO
Figure 6.9 Apparatus Used in Harris' Experiments (Harris, 1974)
POLLINGTONFINEREDWUILDING SANC
80m DPTH
2mr SEAM THICKNESS
1200Wn INITIALPRESSURE
POSITIONSOFMEADINGS .............. +...
SOLID G 0UND LINE ETWEENEXTRACTED
AREAANDSOLID ..........
...... ......
ISOBAIIS .
alt.
e IDE
-.-. .
12
+ 4
. * + +
lq++ +r +* ++
+ + /+ +4 + 1250
*4 4. 4 + . + . ++. + *
+
FMNT AIUTmENTAKA I-.SIODE AMEA

ABUTMENT
Figure 6.10 Vertical Stress Distribution from Harris' Experimental Investigation

(Harris, 1974)
162
I
30"
Plywood base wit.

2.25" round hole
H:and jack wi
ire
aiai gauges
A N1 F.n1-1e 11
transducer
Figure 6.11 Test Apparatus with a Circular Trap Door (Evans, 1984)
163
I
I
I
)
'---~
I
I 1
,---,
I( I
,----
I I I
I
i I I
L,,,J
.L- -l
* I
I i II
I I I
L -,
Trap doors
I
Transducer I
locations 0.56,
0t~Z -H-
(5) I I
00
0\
II V-
I
9.75"
10
=.T
S
I
4.50"
I-
9.75" /
'---- r----,
II SI
j I
I II r SI I
II I .
"--- -- j L_ J ___
-- L J
A. Plan
Plexialas traD door 1...

.se
2"x 2"
timber
B. Profile
Figure 6.12 Test Apparatus with Rectangular Trap Doors (Evans, 1984)
164
050~
"tJ-~~ ONE-r.
I., I..-4.
.r,.ahL
LT
.J:~~L-~:;24 '
____:f~;\~
~ ~
-'~`'
~ .-. -
- lim p,
ft p i--1--
Figure 6.13 Variations of Soil Displacement

ment during an Active Arching
165
0
O
O
O
coN
ooo
-01
0 \0 cos
>
0
>a)
,-p
*
r:)
,1
D ~
0 0
Cr?
ss~aans upifnqzaA0 TpeTT-,
I Jo
7u9oaa
SL
se ooI dBe uo sse.zS TBOT!a9A
166
I
I
1
.
":-,.')
-- . ;.
n ...*
. ... .. .. .
-.: .... ..-
bu~m'",.o. --i.." " ." " .

'
SCAE , "CiSF~c~~'
Figure 6.15 Typical Patterns of Soil Deformation During Passive Arching Tests
(Evans, 1984)
167
0
0 o
'0
cd
oo
0 >
o __
aa
0
0c C.)
I-
0 C 0 0
C\CJ
ssea-S uapfnqq.eAo T'ET!TUI ;Jo 4uaoa

Ssz aooa deal uo ssaa;S T'eOT.;aA
168
Figure 6.17 Soil Deformations Above Trap Doors Lowered in Sequence, Active Arching
Case (Evans, 1984)
169
a
Figure 6.18 Soil Deformations Above Trap Doors Raised in Sequence, Passive Arching
Case (Evans, 1984)
170
-4-
I~I
I I r
I H-
o
4-) M
.r-
0 ;e - H C13
% m 4-) a,
0 0
0
4
CL
a 0d -I 0
HV
H
ul- .1E J
x'
X' ~C/2 'W.-
a)>
0
-N \ II
Co
.C,
-:
0
a
oa -C
0
cn
. ,,
T -L
N o
N
0
0 a.
0
o-
T)o In~e_,__0 \ .
.>
T;'-o - .
--- ----..--- -
o~ " " -- -* - Ge .,
.
a,,
9Z
,2 h o 0o
mr)
o
0 0U
I E 0
SI
w
E/H 0 Tr-,S
171
rJ)
cis
.- o
S(N
o **
C,1
o
cc
Cd
S.r
(O
*i
r- e-1
9/H OTir
172
r*
I .N
C0
oo
o [ ET
(0
0O" m
c,
9L
cq ,--tW4
173
(N
0 -P
0 CI
oc
c.,
0 c
c
u
1,
N1 C
E/H 0Tfll
174
Figure 6.23 The Trap Door Concept (Iglesia et. al., 1990)
175
13"
OCK MASS
MOTOR
GEARBOX
L.UMINUM
'T"D A DT10fD
DISPL
TRAI
Figure 6.24 Centrifuge Trap Door Apparatus (Iglesia et. al., 1990)
176
I
I I
T--Wl l
i
0
V
00
.9u
CI-
aUoO
z
Li
xw
1 00 Cd
O)COC L)
U M
-j
cc ca
0-S W
Gn
(L 0
in
[
c--
i
C4
0
Cu
01 Cu)
E: ".
Cl) -
3oo
HiI
I-s)
ii i
-AL
U.4
(Od/d) 30603 G3ZI1UW1 ONI
177
r
. .. ..'
I 5
I1I
I
I-
I
I
--
I '
"
'- , 1... U3
4
-4
L
0\
0\
a)
a)
(I
c.j4)4
0
W
0, 0 0C,,
c 00
0 v
U)
a)
H H
0
T 00oo
T T I
IT
00
Lq
o
00 =4 =4. 4a)
03 0
C,4
a)
1-I
ZZ ;:z
9.
IOd/d) 30HOA t)ZI1wW:!N
178
- -- I- -~-
-I
4-C~
Nip
Test G1192 - Brick stack
Figure 6.27 Final Configuration of Direct Stack 1/2" Aluminum Rods on 2" Door
(Iglesia et. al., 1990)
179
o2
Figure 6.28 Sketch of Riley's Model Showing the Location of the Hole, and the S,
ric Free Field Point (Riley, 1964)
180
1050 Microseconds
Figure 6.29 Microflash Photographs Showing the Fringe Order Distribution around the
Boundary of the Hole at 1050 Microseconds after Detonation of the Explo-
sive Charge (Riley, 1964)
Figure 6.30 Static and Dynamic Stress Distributions on the Hole Boundary 1050 Micro-
seconds after the Explosion (Riley, 1964)
181
(-
0 S)
0O
Ca.,,.
182
Figure 6.32 Layout of the Shear Box and Sample (Paikowsky et. al., 1995)
183
rotate
Figure 6.33 Plot of the Shear Box and Sample after One Particle Diameter Shear at the
Bottom (Paikowsky et. al., 1995)
184
I
Chapter SEVEN
Summary, Conclusions and Recommendations
7.1 Summary and Conclusions
* General:
(1) Arching is a phenomenon existing widely in nature and man-made construction. It is
an important issue for geotechnical engineers and researchers to comprehend and rec-
ognize in their studies and design.
(2) The arching effect can be defined as follows: "If a portion of an otherwise rigid sup-
port of a geomaterial mass yields, the adjoining soil moves with respect to the remain-
der of the soil mass. This movement is resisted by shearing stresses which reduce the
pressure on the yielding portion of the support while increasing the pressure on the
adjacent rigid portions." Arching can also occur when one portion of a yielding sup-
port moves more than adjoining parts. Generally speaking, arching is stress redistri-
bution occurring in a geomaterial mass.
(3) Arching phenomena can be found in the cavity structures of karst terrain. The collapse
of these cavity structures develop sinkholes.
* Classical Arching Theory:
(4) Silo theory (Janssen, 1895) is the first problem related to arching effect dealt by engi-
neers. In this theory, shear forces between the granular material and the side wall of
the silo played an important part in reducing the vertical stresses on the base of a silo.
185
The vertical normal pressure inside a silo was found to be distributed nonuniformly
across the base.
(5) The first systematic study of the arching mechanism was performed by Terzaghi in
1936 using the trap door tests. Based on these tests, Terzaghi proposed his arching
theory in 1943 where he defined arching as "the transfer of pressure from a yielding
mass of soil onto adjoining stationary parts" (Terzaghi, 1943). He believed that soil
was "arching over" the yielding part of the support. There are several limitations,
however, to Terzaghi's theory: a) the vertical stresses on the horizontal yielding sur-
face are assumed uniform, b) the trap door or yielding strip is assumed rigid, and c) the
sliding surfaces assumed to be vertical in the theory are actually not so.
Analytical Approaches:
(6) Theoretical solutions from continuum approaches using the theory of elasticity can
describe good mechanical behavior of soil only when the soil body has very small de-
formations (Finn, 1963). If one wants to study the arching effect, which means the
geomaterial has a plastic deformation, the continuum approaches by the plasticity the-
ory can predict better mechanical behaviors of soil. Evans (1984) has proven this
point of view in his study.
(7) Theoretical approaches based on discontinuum analysis allow one to study the me-
chanical behavior of granular materials with different particle sizes and shapes (e.g.
Maeda et. al., 1995). However, because of the simplified assumptions and the com-
plexity in the derivation of such theories, their application to discontinuum is still lim-
ited.
186
* Empirical Methods:
(8) Empirical methods for the design of underground openings are either based on very
simple static analyses, or entirely founded on the experience of the proposing person.
In these methods, arching effect is assumed to be the consequence of the local stress
relaxation produced by tunnel excavation. A "ground arch" will transfer most of the
overburden load to the adjacent geomaterials, while only a load from the weight of a
"loosening zone" above the tunnel will act on the tunnel supports.
(9) Due to the uncertainties in empirical methods (e.g. Limited to local experience), these
methods are suggested to be applied in conjunction with analytical methods (e.g. nu-
merical analyses) to confirm their results.
* Experimental Investigations:
(10) Following Terzaghi's original trap door experiment, additional trap door experiments
were performed by others as well. They are similar in principle to Terzaghi's experi-
ment but contain different improvements or extensions. No real advances have been
obtained in these research projects compared to Terzaghi's approach.
(11) The trap door tests of the different researchers show similar results. In active arch-
ing, the trap door is moved away from the soil. The stresses on the trap door decrease
as soon as the trap door is moved, then increase until a steady value is reached. After
the trap is moved, only a small amount of the overburden stresses will act on the trap
door (about 10%). Most of the overburden stresses are carried by the arching. In
187
passive arching, the trap door is moved against the soil body. The stresses on the trap
door increase to a peak value and then decrease slightly to a constant value.
(12) Based on these experiments, three major factors were identified affecting the trap
door test results. They were the overburden depth (H), the ratio of the overburden
depth to the trap door width (H/B), and the relative density of the tested granular soil
(Dr).
(13) Photoelastic methods can be used to explore the interaction and mechanical behavior
of particles in arching.
7.2 Recommendations for Future Research
(1) Various researchers have performed the trap door tests following Terzaghi (1936),
modifying the original trap door tests for different purposes. For example, Harris
(1974) and Evans (1984) installed several trap doors in their tests to study the effects
of an advancing tunnel. Iglesia et. al. (1990) used a centrifuge to study arching effect
in small scale model tests. However, the stress distribution across the yielding surface
and the correct shape of the sliding surfaces are still not known well. For future re-
search, one can try new techniques to measure the distribution of stresses within a soil
body, such as the "Photogrametric method" from Yoshida et. al. (1993). The photo-
grametric method not only can give measurements of the stresses in soil, but also pres-
ents the configuration of soil mass after soil yields. This allows for more detailed
arching mechanism to be explored.
(2) The coefficient of lateral stress (K) was discussed in many trap door studies. Besides
the theoretical method used to predict the force on the trap door from Iglesia et. al.
188
(1990), most of the theoretical approaches and numerical analyses require an appro-
priate estimation of the K value to obtain good results. Hence, more studies about the
coefficient of lateral stress are necessary. Improvement of the measuring techniques
for the horizontal and vertical stresses in soil when yielding takes place, is also re-
quired.
(3) Arching effect is most obvious in granular materials. However, the effects of different
characteristics of individual particles (e.g. particle size, particle shape, etc.) to the
arching formation and stress redistribution in granular materials are still unknown. In
addition, the change of contact forces and displacements of particles within a granular
material are also of great importance. Studies of the interactions between particles are
therefore suggested for future research.
189
References:
Abbott, P.A. (1967)
"Arching for Vertically Buried Prismatic Structures", Journal of the Soil Mechan-
ics and Foundations Division, ASCE, Vol. 93, No. SM5, pp. 233-255.
Balla, A. (1963)
"Rock Pressure Determined from Shearing Resistance", Proc. Int. Conf. Soil Me-
chanics, Budapest, pp. 461.
Beck, B.F. (1984)

"Sinkholes: Their Geology, Engineering and Environmental Impact", Proceedings
of the First Multidisciplinary Conference on Sinkholes, Florida Sinkhole Research
Institute, University of Central Florida, Florida, Orlando, 1984.
Bello, A. A. (1978)
"Simplified Method for Stability Analysis of Underground Openings", Proceed-
ings, First International Symposium on Storage in Excavated Rock Caverns,
Rockstore '77, Stockholm, Sweden, Vol. 2, pp. 289-294.
Benson, R.C. and L.J. LaFountain (1984)

"Evaluation of Subsidence or Collapse Potentials Due to Subsurface Cavities",
Proceedings of the First Multidisciplinary Conference on Sinkholes, Florida Sink-
hole Research Institute, University of Central Florida, Orlando, Florida, pp. 201-
216.
Bierbaumer, A. (1913)
Die Dimensionerung des Tunnelmauerwerks, Engelmann, Leipzig.
Bjerrum, L., C.J. Frimann Clausen, and J.M. Duncan (1972)

"Earth Pressures on Flexible Structures -- A State-of-the-Art Report", Proceed-
ings, Fifth European Conference on Soil Mechanics and Foundation Engineering,
Madrid, Spain, pp. 169-196.
Burghignoli, A. (1981)
"Soil Interaction in Buried Structures", Proceedings, Tenth International Confer-
ence on Soil Mechanics and Foundation Engineering, Stockholm, Sweden, Vol. 2,
pp. 69-74.
Butterfield, R. (1969)
"A Theoretical Study of the Pressures Developed in a Silo Containing single-Sized
Particles in a Regular Packing", International Journal of Rock Mechanics and
Mining Sciences, Pergamon Press, Vol. 6, pp. 227-247.
190
Chelapati, C.V. (1964)
"Arching in Soil Due to the Deflection of a Rigid Horizontal Strip", Proceedings
of the Symposium on Soil-Structure Interaction, University of Arizona, Tucson,
Arizona, pp. 356-377.
Connors, P. (1995)
Examination of Boundary Effects in Interfacial Testing, M.S. Thesis, UMASS,
Lowell.
Davis, R.E. and A.E. Bacher (1968)

"California's Culvert Research Program -- Description, Current Status, and Ob-
served Peripheral Pressures", Highway Research Record, No. 249, pp. 14-23.
Douglas, I. (1965)
"Calcium and Magnesium in Karst Waters", Helictite, Vol. 3, pp. 23-36.
Einstein, H.H., C.W. Schwartz, W. Steiner, M.M. Baligh, and R.E. Levitt (1980)
"Improved Design for Tunnel Supports: Analysis Method and Ground Structure
Behavior: A Review -- Vol. II", MIT, DOT-05-60136.
Evans, C. H. (1983)
An Examination of Arching in Granular Soils, M.S. Thesis, MIT.
Feld, J. (1948)
"Early History and Bibliography of Soil Mechanics", Proceedings, Second Inter-
national Conference on Soil Mechanics and Foundation Engineering, Rotterdam,
Vol. 1, pp. 1-7.
Finn, W.D.L. (1963)

"Boundary Value Problems of Soil Mechanics", Journal of the Soil Mechanics and
Foundation Division, ASCE, Vol. 89, No. SM5,pp. 39-72.
Getzler, Z., M. Gellert, and R. Eitan (1970)

"Analysis of Arching Pressures in Ideal Elastic Soil", Journal of the Soil Mechanics
and Foundations Division, ASCE, Vol. 96, No. SM4, pp. 1357-1372.
Getzler, Z., A. Komornik, and A. Mazurik (1968)

"Model Study on Arching Above Buried Structures", Journal of the Soil Mechan-
ics and Foundations Division, ASCE, Vol. 94, No. SM5,pp. 1123-1141.
Handy, R.L. (1985)

"The Arch in Soil Arching", Journal of Geotechnical Engineering, ASCE, Vol. 3,
No. 3, pp. 302-318.
Harris, G.W. (1974)
191
"A Sandbox Model Used to Examine the Stress Distribution Around a Simulated
Longwall Coal-Face", International Journal of Rock Mechanics, Mining Sciences
and Geomechanical Abstracts, Pergamon Press, Vol. 11, pp. 325-335.
Highway Research Board (1971)

"Structure Analysis and Design of Pipe Culverts", National Cooperative Highway,
Research Report 116.
Iglesia, G., H.H. Einstein, and R.V. Whitman (1990)

"Stochastic and Centrifuge Modeling of Jointed Rock Vol. II - Centrifuge Model-
ing of Jointed Rock", US Air Force Office of Scientific Research.
Jakobson, B. (1958)
"On Pressure in Silos", Proceedings, Conference on Earth Pressure Problems,
Brussels, Vol. 1, pp. 49-54.
Janssen, H.A. (1895)

"Versuche fiber Getreidedruck in Silozellen", Zeitschrift Verein Deutscher Inge-
nieure, Bd XXXIX, pp. 1045-1049.
Jennings, J.N. (1971)

Karst, the MIT Press, pp. 98-143.
Koutsabeloulis, N.C. and D.V. Griffiths (1989)

"Numerical Modeling of the Trapdoor Problem", Geotechnique, Vol. 39, No. 1,
pp. 77-89.
Krynine, D.P. (1945)

"Discussion of Stability and Stiffhess of Cellular Cofferdams by Karl Terzaghi",
Transactions, ASCE, Vol. 110, pp. 1175-1178.
Ladanyi, B. and B. Hoyaux (1969)

"A Study of the Trap-Door Problem in a Granular Mass", Canadian Geotechnical
Journal, Vol. 6, No. 1, pp. 1-15.
LeGrand, H.E. (1973)

"Hydrological and Ecological Problems of Karst Regions", Science, Vol. 179, pp.
859-864.
Luscher, U. and K. H6eg (1964)

"The Beneficial Action of the Surrounding Soil on the Load-Carrying Capacity of
Buried Tubes", Proceedings of the Symposium on Soil-Structure Interaction, Uni-
versity of Arizona, Tucson, Arizona, pp. 393-402.
Luscher, U. and K. H6eg (1965)
192
"The Action of Soil Around Buried Tubes", Proceedings, Sixth International
Conference on Soil Mechanics and Foundation Engineering, Montreal, Canada,
Vol. 2, pp. 396-400.
Maeda, K., K. Miura, and S. Toki (1995)

"Mechanical Properties of Elliptic Microstructure Formed in Granular Materials",
Soils and Foundations, Vol. 35, No. 2, pp. 1-13.
Marston, A. (1930)
"The Theory of External Loads on Closed Conduits in the Light of the Latest Ex-
periments", Bulletin 96, Iowa Engineering Experiment Station, Ames, Iowa.
Matyas, E.L. and J.B. Davis (1983a)

"Prediction of Vertical Earth Loads on Rigid Pipes", Journal of Geotechnical En-
gineering, ASCE, Vol. 109, No. 2, pp. 190-201.
Matyas, E.L. and J.B. Davis (1983b)

"Experimental Study of Earth Loads on Rigid Pipes", Journal of Geotechnical
Engineering, ASCE, Vol. 109, No. 2, pp. 202-209.
McNulty, J.W. (1965)

An Experimental Study of Arching in Sand, Ph.D. Thesis in Civil Engineering,
University of Illinois.
Newton, J.G. (1976)

"Induced Sinkholes -- A Continuing Problem Along Alabama Highways", Interna-
tional Association of Hydrogeological Sciences Proceedings, Anaheim Sympo-
sium, No. 21, pp. 453-463.
Paikowsky, S. G. (1989)
A Static Evaluation of Soil Plug Behavior with Application to the Pile Plugging
Problem, Sc.D. Thesis, MIT, pp. 163-358.
Paikowsky, S.G., K.J. DiRocco, and F. Xi (1993)

"Interparticle Contact Force Analysis and Measurements Using Photoelastic
Techniques", 2nd International Conference on Discrete Element Methods, MIT,
Cambridge, Massachusetts.
Paikowsky, S.G. and F. Xi (1995)

"Kinematics of 2-D Particulate Media Utilizing Image Analysis", 10th ASCE En-
gineering Mechanics Specialty Conference, University of Colorado at Boulder,
Boulder, Colorado.
Paikowsky, S.G., J. Ting, F. Xi, and G.Mischel (1996)
193
"Numerical and Experimental Comparison of Shear Along Granular Material/
Solid Interface", ASME, Mechanics & Material Conference, Johns Hopkins Uni-
versity, Maryland.
Peck, R.B. (1975)

"Lateral Pressures Against Tunnels", Seminar on Lateral Soil Pressures Generated
by Pipes, Piles, Tunnels, and Caissons, Dayton Section ASCE, pp. 14.
Proctor, R.V. and T.L. White (1946)

Rock Tunneling with Steel Supports, Commercial Shearing, Inc.
Proctor, R.V. and T.L. White (1977)

Earth Tunneling with Steel Supports, Commercial Shearing, Inc.
Ranken, R.E. and J. Ghaboussi (1975)

Tunnel Design Considerations: Analysis of Stresses and Deformations Around Ad-
vancing Tunnels, Department of Transportation, Report No. FRA-OR and D 75-
84.
Riley, W.F. (1964)

"Stresses at Tunnel Intersections", Journal of the Engineering Mechanics Division,
ASCE, Vol. 90, No. EM2, pp. 167-179.
Rowe, P.W. (1952)

"Anchored Sheet-Pile Walls", Proceedings, Institute of Civil Engineers, London,
British England, Vol. ,pp. 27-70.
Sakaguchi, H. and E. Ozaki (1992)

"Analysis of the Formation of Arches Plugging the Flow of Granular Materials",
Proceedings of the 2nd International Conference on Discrete Element Method,
MIT, Cambridge, Massachusetts, pp. 153-163.
Selig, E.T. (1975)

"Stresses and Deflections Around Large Corrugated-Metal, Buried Structures",
Seminar on Lateral Soil Pressures Generated by Pipes, Pipes. Tunnels and Cais-
sons, Dayton Section, ASCE, 36 p.
Sowers, G.F. (1984)

"Correction and Protection in Limestone Terrain", Proceedings of the First Multi-
disciplinary Conference on Sinkholes, Florida Sinkhole Research Institute, Uni-
versity of Central Florida, Orlando, Florida, pp. 373-378.
Spangler, M.G. (1964)
194
"Protection of Underground Structures by Arch Action Associated with the Imper-
fect Ditch Method of Construction", Proceedings of the Symposium on Soil-
Structure Interaction University of Arizona, Tucson, Arizona, pp. 531-546.
Spangler, M.G. and R.L. Handy (1973)

"Loads on Underground Conduits", Soil Engineering. 3rd Edition, Harper Collins,
New York, pp. 658-686.
Spangler, M.G. and R.L. Handy (1982)

"Loads on Underground Conduits", Soil Engineering, 4th Edition, Harper Collins,
New York, pp. 727-763.
Stefanoff, G. and B. Boshinov (1977)

"Bearing Capacity of Hollow Piles Driven by Vibration", Proceedings of the 9th
ICSMFE, Tokyo, pp. 753-758.
Stone, K.J.L. (1988)

Modeling of Rupture Development in Soils, Ph.D. Dissertation, Wolfson College,
Cambridge University.
Sz6chy, K. (1966)
The Art of Tunneling, 21st Edition, Akad6miai Kiad6, Budapest.
Szichy, K. (1973)
The Art of Tunneling, 2nd Edition, Akademiai Kiad6, Budapest, pp. 211-243.
Terzaghi, K. (1936)
"Stress Distribution in Dry and in Saturated Sand Above a Yielding Trap-Door",
Proceedings. First International Conference on Soil Mechanics and Foundation
Engineering, Cambridge, Massachusetts, pp. 307-311.
Terzaghi, K. (1943)
Theoretical Soil Mechanics, John Wiley and Sons, New York, pp. 66-76.
Terzaghi, K. and R.B. Peck (1968)

Soil Mechanics in Engineering Practice, 2nd Edition, John Wiley and Sons, New
York, pp. 267-268.
Trollope, D.H. (1957)

"The Systematic Arching Theory Applied to the Stability Analysis of Embank-
ments", Proceedings, Fourth International Conference on Soil Mechanics and
Foundation Engineering, Vol. 2, pp. 382-388.
Trollope, D.H., M.G. Speedie, and I.K. Lee (1963)
195
"Pressure Measurements on Tullaroop Dam Culvert", Forth Australia-New Zea-
land Conference on Soil Mechanics and Foundation Engineering, pp. 81-92.
Trombe, F. (1952)
Trait6 de Sp616ologie, Paris.
Truesdale, W.B. and E. Vey (1964)

"An Investigation of Panel-Arching Effects in Noncohesive Soil", Proceedings of
the Symposium on Soil-Structure Interaction, University of Arizona, Tucson, Ari-
zona, pp. 349-355.
Vardoulakis, I., B. Graf, and G. Gudehus (1981)

"Trap-Door Problem with Dry Sand: A Statical Approach Based Upon Model Test
Kinematics", International Journal for Numerical and Analytical Methods in Geo-
mechanics, John Wiley and Sons, LTD, Vol. 5, pp. 57-78.
Whitman, R.V., Z. Getzler, and K. H6eg (1962)

"Static Tests Upon Thin Domes Buried in Sand", MIT Research Project Report
No. R62-41, December, 1962.
Whitman, R.V., Z. Getzler, and K. H6eg (1963)

"Tests Upon Thin Domes Buried in Sand", Journal of the Boston Society of Civil
Engineers, January, pp. 1-22.
Yoshida, T., F. Tatsuoka, M.S.A. Siddiquee, Y. Kamegai, and C.S. Park (1993)
"Shear Banding in Sands Observed in Plane Strain", Proceedings for the Third In-
ternational Workshop on Localisation and Bifurcation for Soils and Rocks, Aus-
sois, France, September 6-9, 1993.
196

Mit

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Mit

Uploaded by

Copyright:

Available Formats

A Literature Study of the Arching Effect

B.Sc. Civil Engineering (1990)

Submitted to the Department of Civil and Environmental

MASSACHUSETTS INSTITUTE OF TECHNOLOGY

1996 Hsien-Jen Tien

Signature of Author .......... .. e=..........

D blt of Civil and Environmental Engineering

Certified by ............................. ..................

Accepted by .............................. ... r

MASSACHUSETTS INSTITUTE PrOf. Joseph M. Sussman

A LITERATURE STUDY OF THE ARCHING EFFECT

Submitted to the Department of Civil and Environmental Engineering

Thesis Supervisor: Dr. Herbert H. Einstein

The facilities of the Department of Civil and Environmental Engineering in

Finally, I want to thank my parents who always provided support, encouragement,

CHAPTER 1 INTRODUCTION: DEFINITION A

CHAPTER 2 RELEVANCE AND APPLICATIOI

CHAPTER 3 CLASSICAL ARCHING THEORIE

3.2.3 Terzaghi's Arching Theory

CHAPTER 4 ANALYTICAL APPROACHES

CHAPT TER 5 EMPIRICAL METHODS 112

CHAPTER 6 EXPERIMENTAL INVESTIGATIONS AND

CHAPTER 7 SUMMARY, CONCLUSIONS AND RECOMMENDATIONS 185

1.1 General Background

ground structures, for example, underground conduits. Underground openings can be

ability compared to an identical unburied structure. A more detailed definition of arching

derivations, analytical methods, numerical analyses, and experimental investigations. Re-

1.2 Definition of Arching

Arching can be best described as a transfer of forces between a yielding mass of

stresses in the soil.

case at Plane AA and BB are shown in Figure 1.3 (b).

This condition is highly unlikely to be found in natural or man-made environments due to

structure components (like steel or concrete).

Underground structures normally do not have uniform deformation resulting in

and passive arching at the same time.

1.3 Brief History of Arching Studies

The phenomenon of arching has been recognized

generally been sporadic, usually directed toward a parl

point in time. This section summarizes the developments

Arching is present in many geotechnical problem

recognized and investigated in a non-geotechnical co

French military engineers were asked to design magazin

that the base of the silo only supported a fraction of the

small section of the base were detached and lowered, t

experienced was independent of the height of material ir

that an "arch" had formed above this displaced section.

edge of the behavior within magazine silos was utilized

approach to the design of silos for grain and other parl

of the United States. Engineers found that many of the

structural failures subsequent to installation and backfi

Anson Marston (1930) performed extensive research a

1964). and this was attributed to arching.

updated Marston's recommendations in light of experience obtained in the several decades

specific backfilling procedures (Spangler, 1964; Spangler & Handy, 1973).

"Symposium on Soil-Structure Interaction" in 1964.

a good agreement between the simulation outcomes and experimental measurements.

(Paikowsky et. al., 1996).

1.4 Scope of Study

(1) Relevance andApplication of Arching in Soil Mechanics: (Chapter2)

plugging (Paikowsky, 1989).

(2) ClassicalArching Theories: (Chapter 3)

and (d) Soil Plugging (Paikowsky, 1989),.

(3) AnalyticalApproaches: (Chapter 4)